Audio driven facial animation with adaptive speech rate

Abstract

The invention discloses an audio driven facial animation with adaptive speech rate. In various examples, techniques for audio-driven facial animation with adaptive speech include determining that a speech rate associated with an audio segment exceeds a threshold. The technique further includes upsampling a first set of features associated with the audio segment to a second set of features different in size than the first set of features based at least on the speech rate exceeding a threshold. The technique further includes generating a facial animation output corresponding to the audio segment using one or more machine learning models and based at least on at least a subset of the second set of features.

Description

Background Art

[0001] Audio-driven facial animation refers to the use of audio data to drive the generation of animated characters' speech and / or facial expressions that are synchronized with the audio data. Audio-driven facial animation can therefore be used to create realistic, expressive, and engaging characters that appear in games, movies, digital avatar systems, real-time digital assistants, in-car assistants, and / or other environments.

[0002] Audio-driven facial animation is usually performed by analyzing the audio data to extract voice information and then matching the voice information with corresponding facial movements and / or expressions. Traditionally, this process is manually performed by animators with domain expertise via a time-consuming and resource-intensive process. For example, an animator may initially identify different time points at which various phonetic sounds appear in a soundtrack. Next, the animator may generate facial animation synchronized with the soundtrack by manually setting and / or adjusting weights and / or parameters that model the facial expressions and / or movements corresponding to the sounds at these time points. Because a given soundtrack may include a large number of sounds and multiple weights and / or parameters must be set and / or adjusted for each sound, hours of manual work by the animator may be required to produce a few seconds of animation results.

[0003] Recently, machine learning techniques have been employed to automate the process of animating characters based on input audio data. However, these approaches can result in animations that lack sufficient realism and / or expressiveness. For example, a neural network may be trained to animate only the mouth area of ​​a face to match the speech represented by the corresponding audio data. Because the rest of the face remains largely static within the animation, the animation may appear to be dull, emotionless, or unconvincing to a human viewer.

[0004] Although more complex deep learning models are able to generate audio-driven facial animations that animate the entire face and contain the emotions and expressions of the corresponding character, these deep learning models are generally unable to adapt to changes in the speech rate in the input audio. For example, given an input, a deep learning model can output motion or geometric deformation information of different areas of the face (e.g., head, skin, eyes, tongue, etc.), the input including an audio segment (segment) with which the motion or geometric deformation information is to be synchronized, one or more emotions to be presented during the audio segment, and / or other parameters that affect the resulting facial animation. When the audio segment includes a "normal" speech rate (e.g., up to a certain number of phonemes / time period), the deep learning model may be able to generate realistic and expressive facial animations that match the audio segment. However, when the audio segment includes a "fast" speech rate (e.g., more than a certain number of phonemes), the deep learning model may not be able to generate the correct shape for the phonemes found in the segment of the audio. Instead, the deep learning model may generate an overly "smooth" animation and / or a facial shape between two or more facial shapes of two or more phonemes.

[0005] Therefore, more efficient techniques are needed to improve audio-driven facial animation in the presence of variable speech rates. Summary of the invention

[0006] Embodiments of the present disclosure relate to audio-driven facial animation with adaptive speech rate. The techniques described herein include: determining that a rate of speech associated with an audio clip exceeds a threshold. The techniques also include: upsampling a first feature set associated with the audio clip to a second feature set of a different size than the first feature set based at least on the rate of speech exceeding the threshold. The techniques further include: using one or more machine learning models to generate a facial animation output corresponding to the audio clip based at least on at least a subset of the second feature set.

[0007] One technical advantage of the disclosed technology over existing methods is the ability to adapt audio-driven facial animation techniques to audio with varying speech rates. Therefore, when compared to conventional audio-driven facial animation methods that do not account for varying speech rates in the input audio, the disclosed technology can be used to improve the accuracy, realism, and / or expressiveness of facial animation. Another advantage of the disclosed technology is that improvements in animation quality for fast speech rates can be achieved without the need to use audio data that includes varying speech rates to train machine learning models involved in audio-driven facial animation. As a result, the disclosed technology avoids the additional overhead and resource consumption associated with collecting additional training data for the machine learning model and using the additional training data to train the machine learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The present system and method for audio-driven facial animation with adaptive speech rate is described in detail below with reference to the accompanying drawings, wherein:

[0009] Figure 1 A computing device configured to implement one or more aspects of various embodiments is shown;

[0010] Figure 2 A system for performing audio-driven facial animation according to various embodiments is shown, the system comprising Figure 1 Detection engine and processing engine;

[0011] Figure 3 Methods of representing data used in generating animations according to various embodiments are shown;

[0012] Figure 4 According to various embodiments, Figure 1 The detection engine and processing engine process the sample audio input;

[0013] Figure 5 A flow chart showing a method for processing audio input according to various embodiments is shown;

[0014] Figure 6 illustrates components of a distributed system that may be used to generate image or video content according to various embodiments;

[0015] Fig. 7A Inference and / or training logic according to various embodiments is shown;

[0016] Figure 7B Inference and / or training logic according to various embodiments is shown;

[0017] Figure 8 An example data center system is shown in accordance with various embodiments;

[0018] Fig. 9 A computer system according to various embodiments is shown;

[0019] Fig.10 A computer system according to various embodiments is shown;

[0020] Fig.11 illustrates at least a portion of a graphics processor according to one or more embodiments;

[0021] Fig.12 illustrates at least a portion of a graphics processor according to one or more embodiments;

[0022] Fig.13 is an example data flow diagram of a high-level computing pipeline according to at least one embodiment;

[0023] Fig.14 is a system diagram of an example system for training, adapting, instantiating, and deploying machine learning models in an advanced computing pipeline according to at least one embodiment. DETAILED DESCRIPTION

[0024] In the following description, various embodiments will be described. For the purpose of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the embodiments. However, it will also be apparent to those skilled in the art that these embodiments can be practiced without these specific details. In addition, well-known features may be omitted or simplified so as not to obscure the described embodiments.

[0025] The systems and methods described herein may be used by, but are not limited to, non-autonomous vehicles or machines, semi-autonomous vehicles or machines (e.g., in one or more adaptive driver assistance systems (ADAS)), autonomous vehicles or machines, driverless and driverless robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, airships, boats, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, submarines, drones, and / or other vehicle types. Further, the systems and methods described herein may be used for a variety of purposes, including, by way of example and not limitation, for machine control, machine motion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, security and surveillance, simulation and digital twins, autonomous or semi-autonomous machine applications, deep learning, environmental simulation, object or actor simulation and / or digital twins, data center processing, conversational AI, light transport simulation (e.g., ray tracing, path tracing, etc.), collaborative content creation for 3D assets, cloud computing, and / or any other suitable application.

[0026] The disclosed embodiments may be included in a variety of different systems, such as automotive systems (e.g., control systems for autonomous or semi-autonomous machines, perception systems for autonomous or semi-autonomous machines), systems implemented using robots, aerial systems, medical systems, boating systems, smart area monitoring systems, systems for performing deep learning operations, systems for performing simulation operations, systems for performing digital twin operations, systems implemented using edge devices, systems including one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implementing one or more large language models (LLMs), systems implemented at least in part in a data center, systems for performing conversational AI operations, systems for performing light transport simulations, systems for performing collaborative content creation of 3D assets, systems implemented at least in part using cloud computing resources, and / or other types of systems.

[0027] As discussed herein, conventional techniques for performing audio-driven facial animation involve manually setting and / or adjusting weights and / or parameters that represent facial expressions and / or movements of voices at different points in time when different voices appear in an audio track, which is time consuming, resource intensive, and difficult to scale. Although machine learning techniques have been developed to automatically animate characters and / or depict emotions and expressions of corresponding characters based on input audio data, these machine learning techniques cannot adapt to changes in speech rate within the input audio.

[0028] In order to improve the generation of facial animations when there is a varying speech rate in the corresponding input audio track, the disclosed technology determines whether the speech rate within a given audio segment exceeds one or more thresholds. For example, a neural network and / or another type of audio processing component can be used to detect the number of phonemes within a 0.5 second audio segment, the number of transitions between phonemes, and / or another measure of speech rate. The audio processing component can also or alternatively calculate the speech rate as an exponential moving average and / or another aggregation of various measures of speech rates on multiple audio segments falling within a given time window. The speech rate can be compared with one or more thresholds representing different speech rates higher than the normal speech rate. If the detected speech rate meets or exceeds a given threshold, it can be determined that the detected speech rate is at or above the high speech rate represented by the threshold. If the detected speech rate drops below all thresholds, it can be determined that the detected speech rate corresponds to a normal speech rate. The normal speech rate can be a speech rate in a range determined empirically, such as a range of the number of phonemes / time units. Although primarily described with respect to speech rates exceeding thresholds, a similar process may be performed for speech rates below one or more thresholds (eg, speech rates below normal or expected speech rates).

[0029] When it is determined that the speech rate associated with the audio segment exceeds a given threshold, the audio features extracted from the audio segment are upsampled, and some or all of the upsampled audio features are provided as input to one or more machine learning models, which are used to perform facial animation for the audio segment. For example, an encoder and / or another type of neural network can be used to convert the audio segment into a list of not limited to 25 audio features. Within the list, the features can be sorted by time and represent a continuous subset of the audio segment. When the speech rate of the audio segment exceeds a threshold representing a high speech rate, bilinear upsampling can be performed on the audio features to increase the number of audio features by a corresponding factor. A subset of the upsampled audio features can be selected and input into one or more neural networks. Then one or more neural networks can convert the input subset of the upsampled audio features into vertex positions, geometric deformations, and / or other representations of facial shapes or expressions that match the audio segment.

[0030] One technical advantage of the disclosed technology over existing methods is the ability to adapt audio-driven facial animation techniques to audio with varying speech rates. Therefore, when compared to conventional audio-driven facial animation methods that do not account for varying speech rates in the input audio, the disclosed technology can be used to improve the accuracy, realism, and / or expressiveness of facial animation. Another advantage of the disclosed technology is that improvements in animation quality for fast speech rates can be achieved without the need to use audio data that includes varying speech rates to train machine learning models involved in audio-driven facial animation. As a result, the disclosed technology avoids the additional overhead and resource consumption associated with collecting additional training data for the machine learning model and using the additional training data to train the machine learning model.

[0031] Figure 1A computing device 100 configured to implement one or more aspects of various embodiments is shown. In at least one embodiment, the computing device 100 includes a desktop computer, a laptop computer, a smart phone, a personal digital assistant (PDA), a tablet computer, a server, one or more virtual machines, an embedded system, a system on a chip, a computing system of an autonomous, semi-autonomous or non-autonomous machine, a talking kiosk, a smart display, and / or any other type of computing device configured to receive input, process data, and optionally display images, and is suitable for practicing one or more embodiments. The computing device 100 is configured to run a detection engine 122 and a processing engine 124 that may reside in a memory 116. Note that the computing devices described herein are illustrative and any other technically feasible configurations fall within the scope of the present disclosure. For example, multiple instances of the detection engine 122 and / or the processing engine 124 may be executed on a group of nodes in a distributed and / or cloud computing system to implement the functionality of the computing device 100. Alternatively or additionally, the computing device 100 may be configured to communicate with at least about Figure 6-14 The described devices and / or systems can be implemented in a similar manner.

[0032] In at least one embodiment, the computing device 100 includes, but is not limited to, an interconnect (bus) 112 connecting one or more processors 102, an input / output (I / O) device interface 104 coupled to one or more input / output (I / O) devices 108, a memory 116, a storage device 114, and / or a network interface 106. The one or more processors 102 may include any suitable processor implemented as a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), an artificial intelligence (AI) accelerator, a deep learning accelerator (DLA), a parallel processing unit (PPU), a data processing unit (DPU), a vector or visual processing unit (VPU), a programmable visual accelerator (PVA) (which may include one or more VPUs and / or a direct memory access (DMA) system), any other type of processing unit, or a combination of different processing units, such as one or more CPUs configured to operate in conjunction with one or more GPUs. In general, the one or more processors 102 may include any technically feasible hardware unit capable of processing data and / or executing software applications. Further, in the context of the present disclosure, the computing elements shown in computing device 100 may correspond to physical computing systems (e.g., systems in a data center or machine) and / or may correspond to virtual computing instances executing within a computing cloud.

[0033] In at least one embodiment, the I / O devices 108 include devices capable of receiving input, such as a keyboard, a mouse, a touchpad, a VR / MR / AR headset, a gesture recognition system, a steering wheel, a mechanical, digital or touch-sensitive button or input component, and / or a microphone, and devices capable of providing output, such as a display device, a haptic device, and / or a speaker. In addition, the I / O devices 108 may include devices capable of both receiving input and providing output, such as a touch screen, a universal serial bus (USB) port, and the like. The I / O devices 108 may be configured to receive various types of input from an end user (e.g., a designer) of the computing device 100, and also provide different types of output to the end user of the computing device 100, such as a displayed digital image or digital video or text. In some embodiments, one or more I / O devices 108 are configured to couple the computing device 100 to a network 110.

[0034] In at least one embodiment, network 110 is any technically feasible type of communication network that allows data to be exchanged between computing device 100 and an internal, local, remote, or external entity or device (such as a network server or another networked computing device). For example, network 110 may include a wide area network (WAN), a local area network (LAN), a wireless (e.g., WiFi) network, and / or the Internet, among others.

[0035] In at least one embodiment, storage device 114 includes non-volatile storage for applications and data, and may include fixed or removable disk drives, flash memory devices, and CD-ROM, DVD-ROM, Blu-ray, HD-DVD, or other magnetic, optical, or solid-state storage devices. Detection engine 122 and / or processing engine 124 may be stored in storage device 114 and loaded into memory 116 when executed.

[0036] In one embodiment, the memory 116 includes a random access memory (RAM) module, a flash memory unit, and / or any other type of memory unit or combination thereof. The one or more processors 102, the I / O device interface 104, and the network interface 106 may be configured to read data from and write data to the memory 116. The memory 116 may include various software programs, or more specifically, may include software codes that may be executed by the one or more processors 102 and application data associated with the software programs (including the detection engine 122 and / or the processing engine 124).

[0037] The detection engine 122 and the processing engine 124 include functionality for analyzing audio input that is used to drive the generation of animated characters speaking and / or making facial expressions that are synchronized with the audio data. More specifically, the detection engine 122 and the processing engine 124 are configured to adapt a machine learning model used to perform audio-driven facial animation to higher than normal speech rates. As described in further detail herein, the detection engine 122 and the processing engine 124 include functionality for detecting higher than normal speech rates in the audio input and upsampling features generated from the audio input before the features are input into the machine learning model, thereby improving the accuracy, realism, and / or expressiveness of the audio-driven facial animation without the need to train the machine learning model using audio data that includes varying speech rates.

[0038] Figure 2 A system 200 for performing audio-driven facial animation according to various embodiments is shown, the system comprising Figure 1 As mentioned herein, the detection engine 122 and the processing engine 124 operate to detect higher than normal speech rates in the audio input and upsample features generated from the audio input to improve the performance of a machine learning model that uses the audio input and / or the features to perform audio-driven facial animation. Figure 2 As shown, a segment of audio data, such as an audio frame or a segment of audio within a current audio window 202, can be provided as an input to an analysis network 208. The analysis network 208 analyzes the input audio data and converts the input audio data into a feature set 236. For example, the analysis network 208 may include a shared audio decoder and encoder for encoding audio features 236 in the form of an audio feature vector of length 25, which can be provided as an input to an articulation network 210. Within the audio feature vector, the features 236 can be ordered in time, such that features 236 that appear in earlier positions within the audio feature vector represent earlier time periods within the audio window 202, and features 236 that appear in later portions within the audio feature vector represent later time periods within the audio window 202.

[0039] Emotional labels 204 (or vectors) and style vectors 222 may also be provided as inputs to the pronunciation network 210. In some embodiments, only a single vector may be provided as an input including both style and / or emotional data, such as may be fused together using a fusion component 220 or process. The emotional vector 204 may include data for one or more emotions applied to the speech being trained, such as the emotions that a voice actor is instructed to use when uttering the speech captured in the audio data. In some cases, this may include data for a single emotional label, such as "anger," or may include data for multiple emotions, such as "anger" and "sadness," and the potential relative weights of those two emotions. These labels and / or weights may have been provided to the voice actor initially, may have been determined after the speech was spoken, and / or may be included in updated labels after hearing the speech emitted by the audio capture for a particular emotion, as well as other such options.

[0040] In at least one embodiment, the style vector 222 may also be provided as input to the pronunciation network 210 during training (and similarly in deployment and / or inference). The style vector 222 may include data related to any aspect of the animation or movement of a facial component that modifies how one or more points of one or more facial components should move for a given emotion or emotion vector. This may include affecting the movement of a specific facial component, or providing a style of the overall animation to be used, such as "tense" or "professional". The style vector 222 may also be viewed as a finer-grained control of the emotion, where the emotion vector provides one or more labels of one or more emotions for use, and the style provides a finer-grained control of how to express one or more emotions through the animation. Other methods of determining style data may also be used. In different implementations, a single set of emotion and style vectors may be provided for a given audio clip, a set of vectors may be provided for each frame of an animation to be generated, or a set of vectors may be provided for a specific point or frame (e.g., an emotion keyframe) of an animation, where at least one emotion or style value or setting will be modified relative to a previous frame.

[0041] In the system 200, the emotion vector 204 and / or the style vector 222 are fed into the pronunciation network 210 at multiple levels, including at least the beginning and end of the network to help adjust the network. The pronunciation network 210 can use a shared audio encoder and one or more (e.g., multiple) decoders for each facial component (e.g., facial skin, jaw, tongue, eyeball, and head). During training, the output network 212 can generate an output 214 including a set of head / jaw displacements, eyeball rotations, skin / tongue vertex positions, and / or motion vectors (or other motions or deformations) for each feature point of the facial component, whether for each such feature point or only for feature points that have changed relative to the previous frame, as well as other such options. During training, the output 214 can be compared with "true value" data (such as original reconstructed facial data from (e.g., 4D) image capture) to calculate a total loss value. In at least one embodiment, a loss such as an L2 loss can be used for both the position and velocity of the feature points in the output data representation. In at least one embodiment, the loss function used to determine the loss value may include terms for position, motion, and adversarial losses. This loss value may be used during backpropagation to update network parameters (e.g., weights and biases) of the output network 212, the pronunciation network 210, and / or the analysis network 208. Once it is determined that the output network 212, the pronunciation network 210, and / or the analysis network 208 have converged to an acceptable or desired level of accuracy or precision, and / or another end-of-training criterion (e.g., all training data has been processed or a target or maximum number of training iterations has been performed), the trained output network 212, the pronunciation network 210, and / or the analysis network 208 may be provided or deployed for inference.

[0042] After training of the output network 212, the pronunciation network 210, and / or the analysis network 208 is complete, the system 200 may receive audio data (e.g., in some embodiments, only audio data) across the audio window 202 as input, and may infer outputs 214 for various facial components (e.g., head, face, eyeballs, jaw, tongue), which may then be fed to a renderer 216 (e.g., a rendering engine of an animation or video synthesis system) to generate an animation frame 218, which may be one of a series of frames that provide an animation when rendered or played. As discussed in more detail herein, if the generated output 214 is to be modified in some way with respect to how the system 200 will infer vertex positions based on the audio data, then the emotion label 204 and / or style vector 222 data may also be provided as input to the system 200, for example to convey a particular style or facial behavior to be used in the inferred output 214.

[0043] In some embodiments, a generative neural network such as a generative adversarial network (GAN) can be used to directly reason about image data. One or more components of the system 200 can output a vector encoding the position or motion data of each point on a mesh of one or more facial components, and the output vector (or another output, such as a global transformation matrix) can be fed to the renderer 216, which can apply these values ​​to one or more meshes of the character to guide the animation. The output matrix or vector can have a dimension that matches the characteristics of the facial components, such as, for example, but not limited to, 272 facial feature points for skin, 5 facial feature points for head, 5 facial feature points for jaw, 2 facial feature points for eyeballs, 10 facial feature points for tongue (e.g., using PCA compression), etc. Such an approach can provide sufficiently smooth animations so that additional smoothing or post-processing will not be required in at least many cases. However, the system can allow additional smoothing to be applied, such as in the case where the user can specify one or more smoothing parameters.

[0044] In at least one embodiment, an autocorrelation feature can be extracted from the audio data in the audio window 202 for the current frame, and this feature can be fed into the system 200. The size and / or stride of the audio window 202 can be any appropriate or suitable value and can depend in part on the implementation, but can include a time period corresponding to an animation frame for a target frame rate (e.g., 60 Hz) at a minimum, and can include larger windows to consider portions of the audio for nearby frames (e.g., previous or subsequent frames) and / or provide more accurate and smooth animation and more accurate emotion and / or style determination from the input audio. The example system 200 can use one-hot vector encoding to represent different emotions or emotion labels, where the resulting emotion vectors can be concatenated at the beginning and end layers of the U-Net.

[0045] In at least one embodiment, output 214 may include data specifying motion for different facial parts or components. In addition to facial components, output 214 may also include data for other components of the character, and may involve arms, legs, torso, etc. System 200 may also generate accurate motion data for the character for frames or scenes where the character's face is not visible or only partially represented in the scene. For facial parts with non-rigid deformations (such as skin and tongue), principal component analysis (PCA), for example, may be used to compress the captured (e.g., 4D) motion data. This may allow a facial mesh with a large number of points (such as 60,000 points) to be represented by a vector of much smaller dimensions. In such embodiments, the PCA weight vector may be used as a training representation. In one embodiment, a fully connected layer may fuse emotion and / or style data into smaller vectors that may be inserted into various layers of the network as a series.

[0046] For facial parts with rigid transformations (such as the head and teeth), multiple (e.g., but not limited to 5) landmarks or feature points may be selected or identified on the target mesh, and the position increments of these points from the reference position may be used as training representations from which the rigid transformation matrix may be calculated. For the mandible, the five feature points may include points at either end of the mandible, a center point, and two intermediate reference points, where those intermediate points may not be necessary but may help with fine motion control and noise reduction. For rotatable components such as eyeballs, the system 200 may use two rotation values ​​(e.g., pitch and yaw) to represent horizontal and vertical rotations relative to a default orientation. At runtime or during inference, a full 3D facial animation output 214 may be obtained by the system 200. The system 200 may also allow interactive control of the emotion or style of the speech animation by feeding different emotion vectors 204 and / or style vectors 222 into the pronunciation network 210 to allow modification or "fine tuning" of the motion of individual facial components and facial regions, such as may be part of a real-time or near real-time process, or as post-processing.

[0047] In at least one embodiment, the deep neural network may include a U-Net based architecture. The U-Net based architecture may take an audio sequence as input so that the smoothness of the facial animation will be provided by the context of the input. In another embodiment, the network may have an architecture based on a per-frame convolutional neural network (CNN). The per-frame CNN architecture may receive, for example but not limited to, an audio window of approximately 0.5 seconds as input, which may include data from previous and subsequent frames in the sequence, whereby the CNN may predict data from intermediate frames in the sequence. In yet another embodiment, the network may have an architecture based on a recurrent neural network (RNN).

[0048] RNN-based architectures can receive smaller data windows as input and work with shorter delays in real-time or near-real-time applications, such as a delay of only 0.15 seconds, and can predict motion based at least in part on what was previously observed within this window. Other architectures may also be used within the scope of various embodiments, which may also provide animation fluency based at least in part on context provided to or determined by those architectures. Although different architectures can provide competent results, some architectures may perform better in certain situations or for certain aspects of voice or audio-driven facial animation. For example, it was found that U-Net-based architectures are highly accurate for very strong emotions with large deformations or movements, and exhibit very smooth temporal motion. CNN-based architectures perform well for real-time reasoning and can generate very reliable lip sync motion. RNN-based architectures are shown to perform well for real-time or near-real-time applications, have small delays, and exhibit greater relative variability in the above facial movements.

[0049] In at least some embodiments, given an audio window 202, the system 200 is tasked with inferring a facial expression at the center of the audio window 202. The expression can be directly represented as a per-vertex difference vector from a neutral pose in a fixed-topology facial mesh. Once the analysis network 208, the pronunciation network 210, and / or the output network 212 are trained, the mesh can be animated by sliding the audio window 202 over the sound track at a certain step (e.g., a step that reflects the frame rate associated with the facial animation) and independently evaluating the audio window 202 at each time step using the analysis network 208, the pronunciation network 210, and / or the output network 212. Although the analysis network 208, the pronunciation network 210, and / or the output network 212 may not have memory of past animation frames, in practice temporally stable results can be produced.

[0050] An exemplary and non-limiting CNN-based deep neural network consists of one dedicated layer, ten convolutional layers, and two fully connected layers, which can be divided as follows: Figure 2 The analysis network 208, pronunciation network 210 and output network 212 are shown. The data of the input audio window 202 can be fed to the analysis network 208 to produce a sequence of time-varying speech features that will then drive the pronunciation. The analysis network 208 can first extract the raw formant information using fixed function autocorrelation analysis, and then refine it with multiple (e.g., 5) convolutional layers. Through training, the convolutional layers learn to extract short-term features related to facial animation, such as intonation, emphasis, and specific phonemes. Their abstract time-varying representation can be the output of the final convolutional layer.

[0051] The results can be fed to the pronunciation network 210, which includes, but is not limited to, five additional convolutional layers that analyze the temporal evolution of features and ultimately decide on a single abstract feature vector describing the facial pose at the center of the audio window 202. As auxiliary inputs, the pronunciation network 210 accepts (learned) descriptions of emotional state (e.g., emotion labels 204) and / or style (e.g., style vectors 222) to disambiguate between different facial expressions and speaking styles. Emotional state can be represented alone or together with style data as an E-dimensional vector that is directly concatenated to the output of each layer in the pronunciation network, allowing subsequent layers to change their behavior accordingly.

[0052] In at least one embodiment, each layer l outputs F l ×W l ×H l activations, where F l is the number of abstract feature maps, W l is the dimension of the time axis, and H l is the dimension of the resonance peak axis. In the analysis network 208, a convolution with a stride of 1×3 can be used to gradually reduce H l , while increasing F l , for example, to push the original formant information to the abstract features until the final H l =1 and F l =256. Similarly, 3×1 convolutions can be used in the pronunciation network 210 to reduce W l , for example, to downsample the time axis by combining information from temporal neighbors. Specific parameters can be chosen that consistently perform well while resulting in reasonable training times. The results are not very sensitive to the exact number of layers or feature maps, but it may be beneficial in some systems to organize the convolutions in two different stages to avoid overfitting. The analysis network 208 may perform the same operation at each point along the time axis in order to benefit from the same training samples at different time offsets.

[0053] The example pronunciation network 210 can output a set of 256+E+S abstract features that together represent the desired facial pose - for example, E is used for the dimension of the emotion vector and S is used for the dimension of the style vector. These features can be fed to the output network 212 to produce an animation output 214, which includes the final 3D positions of the set of control vertices in the tracking mesh. The output network 212 can be implemented as a pair of fully connected layers that perform a simple linear transformation on the data. The first layer maps the input feature set to the weights of the linear basis, and the second layer's set calculates the final PCA coefficients for the face and tongue, the rotation values ​​for the eyeballs, and the translation displacements of the jaw and head. The second layer can be initialized to, for example, 150 pre-computed PCA components, which together explain approximately 99.9% of the variation seen in the training data.

[0054] The main input of system 200 may include a speech audio signal, which may be converted into a format such as 16kHz mono audio. The volume of each channel may be normalized to use the full [-1, +1] dynamic range, but such a system may or may not employ other types of processing, such as dynamic range compression, noise reduction, or pre-emphasis filters. An autocorrelation layer may convert an input audio window into a compact 2D representation for subsequent convolutional layers. The resonant frequencies (formants) of the linear filters may carry the necessary information about the phoneme content of the speech. The excitation signal indicates the pitch, timbre, and other characteristics of the speaker's voice, which may not be so important for facial animation, so that the system may focus on the formants to improve generalization to different speakers at least to some extent.

[0055] A method for performing source-filter separation can be based at least in part on linear predictive coding (LPC). LPC decomposes the signal into several short frames, solves the coefficients of the linear filter for each frame based on the first K autocorrelation coefficients, and performs inverse filtering to extract the excitation signal. The resonant frequency of the filter is completely determined by the autocorrelation coefficients, at least some of the processing steps can be skipped, and the system can directly use the autocorrelation coefficients as a representation of the instantaneous resonance peak information. This representation can be suitable for convolutional networks because the layers can easily learn to estimate the instantaneous power of a specific frequency band.

[0056] In one example implementation, 520ms worth of audio is used as input, e.g., 260ms of past and future samples about the desired output pose. This value is selected to capture correlation effects, such as phoneme co-articulation, without providing too much data that may cause overfitting. The input audio window 202 can be divided into 64 audio frames with 2x overlap, so that each frame corresponds to 16ms (256 samples), and consecutive frames are positioned 8ms (128 samples) apart. For each audio frame, the DC component can be removed and a Hann window can be applied to reduce the effect of temporal aliasing. Finally, K=32 autocorrelation coefficients can be calculated to produce a total of 64×32 scalars for the input audio window 202. Although much fewer autocorrelations (e.g., K=12) are sufficient to identify individual phonemes, a method may choose to retain more information about the original signal to allow subsequent layers to also detect changes in pitch.

[0057] Inferring facial animation from speech can be an inherently ambiguous problem, since the same sound can be produced with very different facial expressions. This is especially true for eyes and eyebrows, which have no direct causal relationship with sound production. Such ambiguity is also problematic for deep neural networks, since the training data will inevitably contain cases where nearly identical audio inputs are expected to produce very different output gestures. If the network has nothing else to do except the audio, it will learn to output the statistical mean of conflicting outputs.

[0058] An example method for resolving such ambiguities is to introduce at least an auxiliary input. A small amount of additional potential data may be associated with each training sample to provide enough information for the analysis network 208, the pronunciation network 210, and / or the output network 212, thereby inferring the correct output gesture without ambiguity. This additional data may encode all relevant aspects of the animation in the neighborhood of a given training sample, which cannot be inferred from the audio itself, including different facial expressions and co-articulation patterns. The auxiliary input may include predefined labels and may at least represent the emotional state of an actor. In addition to resolving ambiguities in the training data, such an auxiliary input may also be very useful for reasoning, because it enables the system to mix and match different emotional states with a given channel to provide powerful control over the resulting animation.

[0059] In addition to or alternatively to relying on predefined labels, the system 200 can also adopt a data-driven approach, in which the analysis network 208, the pronunciation network 210, and / or the output network 212 automatically learn a concise representation of the style as part of the training process. This allows the system 200 to extract meaningful emotional states even from in-character footage, as long as there is a sufficient range of emotions. In at least one embodiment, the style state can be represented by an S-dimensional style vector 222, where S is an adjustable parameter that can be set to values ​​such as, but not limited to, 16 or 24, and the components are initialized to random values ​​drawn from a Gaussian distribution. One such vector can be assigned to each training sample, and the matrix storing these latent variables is referred to herein as a style database. The style data can be attached to a list of activations of all layers of the pronunciation network 210, which can make it part of the computational graph of the loss function, and as a trainable parameter, it can be updated during backpropagation with the network weights. In this example, the dimensionality of S is a compromise between two effects. If S is too low, the style cannot disambiguate changes in the training data, resulting in a weak audio response. If S is too high, the style may become too technical to be useful for general reasoning.

[0060] By design, the information provided by the input audio can be limited to short-term effects within, for example, 520ms intervals. As a result, a natural way to prevent styles from containing repetitive information is to prohibit them from containing short-term changes. Focusing styles on long-term effects may also be desirable for inference, as the network may be expected to produce plausible animations even when the emotional state is held fixed. This requirement can be expressed by introducing a dedicated regularization term in the loss function to penalize fast changes in the style database, which results in incremental smoothing of the emotional state during training. One potential limitation to such an approach is that aspects such as blinks and eye movements may not be modeled correctly, as they are not correlated with the audio and cannot be represented using slowly changing emotional states.

[0061] In some embodiments, emotion and style states can be attached to all layers of the pronunciation network 210 to help significantly improve results in practice, because emotion and style states can control animations at multiple levels of abstraction, and higher levels of abstraction can be more difficult to learn. Connecting to earlier layers provides subtle control over subtle animation features (such as co-articulation), while connecting to later layers provides more direct control over output gestures. Early stages of training can focus on the latter, while later stages can focus on the former once individual gestures are reasonably well represented.

[0062] In one method of training the analysis network 208, the pronunciation network 210, and / or the output network 212, an unstructured mesh with texture and optical flow data can be reconstructed from, for example, nine images captured for each frame. A fixed topology template mesh created using a separate photogrammetry pipeline prior to the capture work can be projected onto the unstructured mesh and associated with the optical flow. The template mesh can be tracked across the performance, and any problems are semi-automatically fixed, such as by a tracking artist in software. Several key vertices of the tracking mesh can be used to stabilize the position and orientation of the head. Finally, the vertex positions of the mesh can be derived for each frame in the shot. These positions (or more precisely, deltas from a neutral pose) can be the target output of the network when given an audio window during training.

[0063] For each actor, the training set can be composed of at least two parts: a pangram and in-role material. In general, the quality of reasoning can be improved as the training set grows, but due to the cost of capturing high-quality training data, a small training set may be highly desirable. In at least one embodiment, it is empirically determined that approximately 3 to 5 minutes per actor represents an actual sweet spot. The pangram set can attempt to cover a set of possible facial movements during normal speech in a given target language (such as English). The actor speaks one to three pangrams (e.g., sentences designed to contain as many different phonemes as possible) in several different emotional tones to provide good coverage of the range of expressions. The in-role material set can take advantage of the fact that the performance of the actor's role is usually seriously deviated in terms of emotional and expressive range due to different dramatic and narrative reasons. In the case of a film or game product, the material can be composed of a preliminary version of the script. Only shots that are considered to support different aspects of the role can be selected to ensure that the trained network produces outputs that stay in the role, even if the reasoning is not perfect, or even if a completely novel or voice action that does not conform to the role is encountered.

[0064] Given the potentially ambiguous nature of the training data, efforts can be made to define a meaningful loss function to be optimized. In at least one embodiment, a specialized loss function consisting of three different terms can be used: a position term to ensure that the overall position of each output vertex is approximately correct; a motion term to ensure that the vertices exhibit the right kind of motion under animation; and a regularization term to discourage the style database from containing short-term variations.

[0065] In practice, simultaneous optimization of multiple loss terms can be difficult because these terms can have wildly different magnitudes during training and their balance can change in unpredictable ways. One solution is to associate predefined weights with each term to ensure that the optimization does not overlook them. However, selecting optimal values ​​for the weights can be a tedious process of trial and error, which may need to be repeated every time the training set changes. To overcome these problems, a normalization scheme can be used that automatically balances the loss terms with respect to their relative importance. Thus, an equal amount of effort can be invested in optimizing each term so that no additional weights need to be specified.

[0066] One error metric that can be used is the difference between the desired output y and the output 214 produced by the output network 212. For a given training sample x, this can be expressed using the position term P(x):

[0067]

[0068] Here, V represents the total number of output features including skin / tongue PCA coefficients, eyeball rotation values, and jaw / head translation displacements, and y ( i ) It means y=(y (1) ,y (2) ,...,y (3V) ). In this example, the total number of output components is 3V, because the network outputs the full 3D position of each vertex. Even if the position term ensures that the output of the network is approximately correct at any given moment, it may not be sufficient to produce high-quality animations in all cases. It is observed that training a network with only the position term can result in a large amount of temporal instability, and the response to individual phonemes is generally weak. Therefore, the analysis network 208, the pronunciation network 210 and / or the output network 212 can also be optimized with respect to vertex motion: a given output vertex should move only if it also moves in the training data, and it only moves at the right time. The system 200 can therefore account for vertex motion as part of the loss function.

[0069] One approach for training a neural network is to iterate over the training data in minibatches, where each minibatch consists of B randomly selected training examples x1, x2, ..., x B To account for vertex motion, samples can be plotted as B / 2 time pairs, each consisting of two adjacent frames. The operator m[·] can be defined as the finite difference between pairs of frames, which allows the motion term M(x) to be defined as:

[0070]

[0071] In this equation, a factor of 2 appears because M is evaluated once per time pair.

[0072] Additionally, it would be beneficial to ensure that the network correctly attributes short-term effects to the audio signal and long-term effects to the emotional state. One approach could be to define the regularization term for the emotion / style database using the same finite difference operator as above:

[0073]

[0074] Here, e( i )(x) represents the i-th component of the training sample x stored by the emotion database. It can be noted that this definition does not explicitly prohibit the emotion / style database from containing short-term changes - but rather discourages excessive changes in the average. This may be significant in at least some instances because the training data may occasionally contain legitimate short-term changes in emotional state, and it is undesirable for the analysis network 208, the pronunciation network 210, and / or the output network 212 to incorrectly attempt to interpret them based on the audio signal.

[0075] The caveat of Equation 3 is that R'(x) can be made arbitrarily close to zero by simply decreasing the magnitude of e(x) while increasing the corresponding weights in the analysis network 208, the pronunciation network 210, and / or the output network 212. To illustrate the idea of ​​batch normalization, one can normalize the values ​​relative to the observed magnitude of e(x) by To remove this trivial solution:

[0076]

[0077] To balance these three loss terms, one approach is to exploit the properties of the Adam (or other) optimization method used to train the analysis network 208, the pronunciation network 210, and / or the output network 212. In practice, Adam updates the weights of the network according to the gradient of the loss function, which is normalized in a component-by-component manner according to its long-term estimate of the second primitive moment. This normalization makes the training robust to differences in the magnitude of the loss function, but this is only true for the loss function as a whole - not for individual terms. One approach is to perform similar normalization for each loss term separately. Using the position term as an example, the second primitive moment of P(x) can be estimated for each mini-batch and a moving average v is maintained across consecutive mini-batches. P t , as can be given by the following formula:

[0078]

[0079] Here, t represents the mini-batch index and β is a decay parameter of the moving average that can be set to a value such as, but not limited to, 0.99. The system can be initialized and correct this estimate to take the startup bias into account to obtain Then we can calculate the mean value P(x) on the current mini-batch and use Normalize the value:

[0080]

[0081] In Equation 6, ∈ can be set to, for example, 10- 8 Repeating Equations 5 and 6 for M(·) and R(·), the final loss function can be expressed as In some embodiments, the importance of the loss term can be further fine-tuned by adding weights.

[0082] In at least one embodiment, random time shifts can be utilized to train samples to improve temporal stability and reduce overfitting. Whenever a small batch is presented to the analysis network 208, the pronunciation network 210, and / or the output network 212, the input audio window can be randomly shifted up to 16.6ms (± 0.5 frames at 30FPS) in either direction. In order to compensate, the same shift can be applied by linear interpolation for the desired output posture. The two training samples in the time pair can be shifted by the same amount, wherein different random shift amounts are used for different pairs. In some embodiments, instead of linear interpolation or in addition to linear interpolation, the output of cubic interpolation can also be used.

[0083] To improve generalization and avoid overfitting, multiplicative noise can be applied to the inputs of individual convolutional layers. The noise can have the same magnitude for each layer and can be applied on a per-feature-map basis so that all activations for a given feature map are multiplied by the same factor. The same noise can be applied to pairs of training examples to obtain meaningful motion terms. One formula for this noise is 1.4 N(0,1) There may be no other types of noise or enhancement applied to the training samples other than time shifting of the input / output and multiplicative noise within the analysis network 208, pronunciation network 210, and / or output network 212. However, some methods may perform operations such as adjusting volume, adding reverberation (both long and short), performing time stretching and pitch shifting, applying nonlinear distortion, random equalization, and scrambling phase information, among other such options.

[0084] Once trained, the analysis network 208, pronunciation network 210 and / or output network 212 can be evaluated at any point in time by selecting an appropriate audio window 202 and / or stride, resulting in facial animation at a desired frame rate. The latency of such an approach can depend at least in part on the audio window size, which can extend time periods into the past and / or future. Co-articulation can set a lower limit for look-ahead; it has been observed that look-ahead can be limited to values ​​such as 100ms with little degradation in quality during training, even though some co-articulation effects may be longer. Shortening the look-ahead further than this can in some cases result in a rapid drop in perceived responsiveness, so a practical lower limit for the latency of one embodiment can be set to about 100ms.

[0085] When inferring facial gestures for novel audio, the analysis network 208, the pronunciation network 210, and / or the output network 212 can be provided with the emotional state vector 204 and / or the style vector 222 as auxiliary inputs, which can also be part of a single emotion vector. As part of training, one or more of these networks can learn a vector (e.g., a latent E-dimensional vector) for each training sample, and the emotion database can be used to obtain a robust emotion vector that can be used during inference.

[0086] During training, the analysis network 208, the pronunciation network 210, and / or the output network 212 may attempt to separate latent information (e.g., everything that cannot be inferred from the audio alone) into the emotion / style database. However, this decomposition may result in some amount of crosstalk between pronunciation and overall expression. In practice, many of the learned emotion / style vectors may only apply to the neighborhood of their corresponding training frames and are not necessarily used for general reasoning. In at least one embodiment, the process may use a three-step process to mine robust emotion / style vectors. A problem experienced in many learned emotion vectors is that they de-emphasize mouth movement: when such vectors are used as constant inputs when performing reasoning on new audio, the apparent movement of the mouth may be suppressed. One approach is to pick a few audio windows from a validation set containing bilabial sounds and a few audio windows from a validation set containing vowels for which the mouth should be closed and open, respectively. The emotion / style database may then be browsed to find vectors that exhibit the desired behavior for all selected windows. This preliminary elimination is performed for character 1 obtained from the 100 candidate emotion vectors for further consideration, and this response may vary with different emotion vectors.

[0087] In this example, the second operation in the elimination process is used to play the verification audio track and check the facial movements inferred with each of the candidate emotion / style vectors. At this stage, vectors can be discarded, which results in relaxed or false, unnatural movements, indicating that the vector may be contaminated by short-term effects. This stage narrows the setting to 86 candidate emotion vectors for character 1. As the third and final operation in this example, reasoning can be run on several seconds of audio from different speakers and vectors that eliminate silence or unnatural responses. For character 1, this step leaves 33 emotion vectors.

[0088] The resulting output 214 can be examined for several novel audio clips with each remaining emotion / style vector and a semantic meaning (e.g., "neutral," "pleasant," "surprised," etc.) assigned to each of the several novel audio clips, which depends at least in part on factors such as the emotional state they convey. Which semantic emotions are retained can depend on the training material, and if the training data does not contain such material that is sufficient to generalize to novel audio, it may be impossible to extract, for example, a "happy" emotion. Even after removing all emotion vectors except the best performing emotion vector, there may still be substantial changes to choose from. It is observed that the emotion vectors mined in this way perform well under interpolation, such as sweeping from one emotion vector to another tends to produce results that look natural. Therefore, it is possible to change the emotional state during reasoning based on high-level information from the game engine or by manually finding key frames.

[0089] The resulting facial animation can be highly stable. The main sources of this temporal stability can include motion items and time-shift enhancement, but even with these techniques, a small amount of jitter may remain, for example in the lip region for some inputs at a 4 ms time scale. This may be caused by aliasing between adjacent audio frames around features such as stops and plosives. This may be mitigated at least in part using at least some amount of integration: the analysis network 208, the articulation network 210, and / or the output network 212 are evaluated twice for a given animation frame interval time (e.g., 4 ms), and the predictions are averaged.

[0090] As mentioned above, this approach can also support retargeting. When training the model, the output network 212 can become dedicated to a specific grid. For many operations, it may be desirable to use audio input to drive several different grids. The methods discussed herein can support retargeting of morphs, or transfer of morphing behavior between characters, or transfer of the same character for different stages in life, and other such options.

[0091] like Figure 2As shown, the detection engine 122 determines a speech rate 232 associated with the audio input within a given audio window 202. For example, the detection engine 122 may use a neural network and / or another type of audio processing component to detect the number of phonemes, the number of transitions between phonemes, and / or another measure of the speech rate 232 within an audio segment spanning a 0.5 second audio window 202. The detection engine 122 may also or alternatively calculate the speech rate 232 as an exponential moving average and / or another aggregate of various measures of speech rate (e.g., the number of phonemes, the number of transitions between phonemes, etc.) over one or more audio windows spanning a given time interval within the audio input.

[0092] The detection engine 122 compares the detected speech rate 232 to one or more thresholds 234 representing various speech rates that are above a (e.g., predefined) normal speech rate. If the detected speech rate 232 meets or exceeds a given threshold, then it can be determined that the detected speech rate is at or above a high speech rate represented by the threshold. If the detected speech rate drops below all thresholds 234, then it can be determined that the detected speech rate corresponds to a normal speech rate.

[0093] When the detection engine 122 determines that the speech rate 232 for a given audio segment falls below all thresholds 234, the detection engine 122 passes the features 236 generated by the analysis network 208 from the audio segment to the pronunciation network 210. As discussed herein, the pronunciation network 210 processes the passed features 236 at multiple levels along with the emotion tags 204 and / or style vectors 222. The pronunciation network 210 generates abstract features representing the desired facial pose, which are fed to the output network 212 to produce an animated output 214 corresponding to the audio segment.

[0094] When it is determined that the speech rate within a given audio segment exceeds a given threshold, the detection engine 122 transmits the features 236 generated by the analysis network 208 from the audio segment to the processing engine 124. The processing engine 124 converts the features 236 into a set of corresponding upsampled features 238, and extracts a subset of the upsampled features 238 as a set of pruned features 240.

[0095] For example, the processing engine 124 may perform bilinear upsampling on the 25 features 236 to generate upsampled features 238 that are greater in number by a factor or multiple than the features 236. The processing engine 124 may also extract trimmed features 240 as 25 continuous features (or a different number of continuous features) from the temporal “center” of the upsampled features 238 (e.g., a subset of the upsampled features 238 that are flanked on either side by substantially the same amount of additional upsampled features 238). The processing engine 124 may then provide the trimmed features 240 as input to the pronunciation network 210 for additional processing by the pronunciation network 210 and the output network 212. Because the trimmed features 240 generated by the processing engine 124 are equal in number to the original feature set 236 output by the analysis network 208, these trimmed features 240 may be input into the pronunciation network 210 without requiring additional modifications to the pronunciation network 210.

[0096] When the detection engine 122 applies multiple thresholds 234 to the speech rate 232 of a given audio segment, the processing engine 124 can generate upsampled features 238 and / or cropped features 240 based on the highest threshold met or exceeded by the audio segment. For example, the thresholds 234 can include a first threshold of a speech rate 1.5 times higher than a "normal" speech rate, a second threshold of a speech rate twice the normal speech rate, and a third threshold of a speech rate three times the normal speech rate. The first, second, and third thresholds 234 can be associated with upsampling factors of 1.5, 2, and 3, respectively, to account for the differences in speech rates represented by these thresholds 234. When the audio segment only exceeds the first threshold, the processing engine 124 can use the upsampling factor of 1.5 for the first threshold to convert the features 236 generated by the analysis network 208 into 37 or 38 upsampled features 238. The processing engine 124 may then generate cropped features 240 by extracting 25 upsampled features 238 from the temporal center of these upsampled features 238 (eg, extracting the 7th element to the 31st element from the vector of 37 upsampled features 238 ).

[0097] Continuing with the above example, when the audio segment exceeds both the first and second thresholds 234 but does not exceed the third threshold, the processing engine 124 may use an upsampling factor of 2 for the second threshold to convert the 25 features 236 generated by the analysis network 208 into 50 upsampled features 238. The processing engine 124 may then generate the pruned features 240 by extracting the 25 upsampled features 238 from the temporal center of these upsampled features 238 (e.g., extracting the 13th to 37th elements from the vector of the 50 upsampled features 238).

[0098] Continuing with the above example, when the audio segment exceeds all three thresholds 234, the processing engine 124 can use an upsampling factor of 3 for the third threshold to convert the 25 features 236 generated by the analysis network 208 into 75 upsampled features 238. The processing engine 124 can then generate the pruned features 240 by extracting the 25 upsampled features 238 from the temporal center of these upsampled features 238 (e.g., extracting the 26th element to the 51st element from the vector of 75 upsampled features 238).

[0099] After the processing engine 124 generates the cropped features 240, the processing engine 124 inputs the cropped features 240 into the pronunciation network 210 as a replacement for the set of initial features 236 generated by the analysis network 208. When the cropped features 240 are provided as input to the pronunciation network 210 instead of the features 236, the detection engine 122 and the processing engine 124 allow the pronunciation network 210 and the output network 212 to generate outputs 214 corresponding to phonemes found within the central portion of the audio segment. Thus, when the speech rate 232 indicates multiple phonemes within the audio window 202, the detection engine 122 and the processing engine 124 can avoid generating overly “smooth” animations for two or more phonemes and / or facial shapes between two or more facial shapes.

[0100] Figure 3 An example input vector 300 including audio and an emotion vector that can be provided as input in at least one embodiment is shown. In this example, an audio input 302 can be received, including audio data over at least a time window. The audio input 302 can be used to generate an audio vector from features extracted from the audio. As described above, this can include performing a process that can generate an audio vector such as Figure 3The audio autocorrelation of the audio feature matrix 304 shown is extracted, and these values ​​can be used to generate audio vectors for input to a deep neural network. Similarly, a set of emotional labels can be presented, each of which can be represented as a vector-so that one emotional label can be provided to replace another emotional label in the example presented or described herein. In this example, the emotional vector 310 can indicate which emotion is conveyed (or conveyed for training purposes) in the audio frame, where here, a single emotion is indicated by setting the value of the feature in the vector to 1 and setting the values ​​of all other emotional features to 0. Various other values ​​or sets of values ​​can also be used in other embodiments. In some instances, the emotional vector can be a mixed vector 320 representing a mixed emotional state. In this example, the mixed vector includes a value of 0.65 for the emotion "joy" and a value of 0.35 for the emotion "sadness". Here, in the case where the values ​​add up to 1, in order to represent a 100% emotional state, this indicates that the emotional state is 65% joy and 35% sadness. As discussed elsewhere herein, these weights need not add up to 100%, but instead, each emotion may be set to a value between 0% and 100%, or another such value. Thus, a sadness value of 100% and a joy value of 100% in one system may be equal to a sadness value of 0.5 and a joy value of 0.5, as well as other such options, to indicate an equal amount of each emotion contributing to the overall emotional state.

[0101] In some embodiments, style vectors may be used to encode residual facial expressions that are not clearly represented by the audio input and emotion labels. For example, a style vector may include values ​​to be applied (directly or indirectly) to specific facial components or feature points, such as to cause a character to lift his or her upper lip more in one direction when speaking. The neural network may also learn specific style features during the training process to use for the style vector, as well as other such options, such as those discussed in more detail later in this document.

[0102] Figure 4 It shows the Figure 1 The detection engine 122 and the processing engine 124 process the example audio input 402. Figure 4 As shown, the detection engine 122 processes the audio input 402 across a given audio window to determine a phoneme sequence 404 within the audio input 402. For example, the detection engine 122 may execute a deep neural network trained to perform phoneme detection to extract the phonemes 404 "k", "m", and "i" from the audio input 402.

[0103] The detection engine 122 also uses the number of phonemes 404 detected in the audio input 402 as an indication of the speech rate 232 and applies one or more thresholds 234 to the speech rate 232. For example, the detection engine 122 can use a numeric threshold 234 for the number of phonemes within an audio window to determine whether the number of phonemes 404 detected in the audio input 402 indicates a “higher than normal” speech rate 232. As discussed herein, the detection engine 122 can also or alternatively use a different metric (e.g., the number of transitions between phonemes) and / or a combination of metrics with corresponding thresholds 234 to determine whether the audio input 402 includes a higher than normal speech rate 232.

[0104] exist Figure 4 In the example of , the detection engine 122 determines that the number of phonemes 404 detected in the audio input 402 exceeds one or more thresholds 234. The detection engine 122 also determines an upsampling factor 406 corresponding to the maximum threshold value exceeded by the number of detected phonemes 404. The detection engine 122 also transmits the set of features 236 extracted from the audio input 402 by the analysis network 208 and the upsampling factor 406 to the processing engine 124.

[0105] The processing engine 124 upsamples the features 236 by the upsampling factor 406 to generate a corresponding set of upsampled features 238. For example, the processing engine 124 may perform bilinear upsampling on the features 236 using a value of 2 for the upsampling factor 406 to convert 25 features 236 into 50 upsampled features 238. During this bilinear upsampling, the processing engine 124 may convert a first audio feature vector including 7 features 236 associated with phoneme “k”, 7 features 236 associated with phoneme “m”, and 11 features associated with phoneme “i” into a second audio feature vector including 14 features 236 associated with phoneme “k”, 14 features 236 associated with phoneme “m”, and 22 features associated with phoneme “i”.

[0106] The processing engine 124 then generates the pruned features 240 by extracting a subset of the temporally centered upsampled features 238. Continuing with the above example, the processing engine 124 may generate the pruned features 240 by extracting the last 2 features 236 associated with the phoneme "k", all 14 features 236 associated with the phoneme "m", and the first 11 features associated with the phoneme "i" from the upsampled features 238.

[0107] The processing engine 124 can then provide the tailored features 238 as input to the pronunciation network 210. Because most of the tailored features 238 are associated with the middle phoneme "m", the pronunciation network 210 and the output network 212 can be able to convert the tailored features 238 into more accurate facial animations than if a set of original features 236 associated with all three phonemes 404 were provided as input to the pronunciation network 210.

[0108] See now Figure 5 , each block of the method 500 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, different functions can be implemented by a processor executing instructions stored in a memory. The method can also be implemented as computer-usable instructions stored on a computer storage medium. To name a few, the method can be provided by a stand-alone application, a service or a hosted service (standalone or in combination with other hosted services), or a plug-in for another product. In addition, by way of example, Figure 1-2 The method 500 is described with reference to a system. However, these methods may be additionally or alternatively performed by any system or any combination of systems, including but not limited to those described herein. Further, the operations in the method 500 may be omitted, repeated, and / or performed in any order without departing from the scope of the present disclosure.

[0109] Figure 5 1 shows a flow chart of a method 500 for processing audio input according to various embodiments. Figure 5 As shown, the method 500 starts at operation 502, where the system 200 extracts an audio segment spanning an audio window from an audio sample. For example, the system 200 may extract an audio segment of 0.5 seconds from an audio sample.

[0110] In operation 504, the analysis network 208 converts the audio segment into a time-varying feature sequence. Continuing with the above example, the analysis network 208 may extract raw formant information using fixed function autocorrelation analysis, and then refine the formant information using multiple (eg, 5) convolutional layers.

[0111] In operation 506, the detection engine 122 determines a speech rate associated with the audio segment spanning the audio window. For example, the detection engine 122 may use a deep neural network, a signal processing technique, and / or another technique to detect the speech rate associated with the audio segment as the number of phonemes, the number of transitions between phonemes, and / or another measure of speech rate within the audio segment.

[0112] In operation 508, the detection engine 122 determines whether the speech rate exceeds one or more thresholds. For example, the detection engine 122 can calculate an exponential moving average of the speech rate over a series of audio windows that span a certain time interval within the audio segment and end with the audio window used to determine the speech rate in operation 506. The detection engine 122 can also compare the exponential moving average of the speech rate of the corresponding audio windows (including the audio window used to determine the speech rate in operation 506) and / or one or more individual metrics of the speech rate with one or more digital thresholds representing individual speech rates that are higher than normal speech rates.

[0113] When the detection engine 122 determines that the speech rate does not exceed any threshold, the pronunciation network 210 performs operation 510, in which the pronunciation network 210 converts the features generated in operation 504 into an additional feature set together with the emotion and / or style vectors. For example, the pronunciation network 210 may include five convolutional layers that analyze the temporal evolution of the features to generate an abstract feature vector describing the facial gesture at the center of the audio window. As auxiliary inputs, the pronunciation network may accept (learned) descriptions of emotional state and / or style to disambiguate between different facial expressions and speaking styles. The emotional state may be represented alone or together with the style data as an E-dimensional vector that is directly concatenated to the output of each layer in the pronunciation network, enabling subsequent layers to change their behavior accordingly.

[0114] In operation 512, the output network 212 converts the additional feature set from the pronunciation network 210 into a facial animation output for the audio clip. For example, the output network 212 may generate an output including, but not limited to, head displacement, jaw displacement, eye rotation, a set of skin vertex positions, a set of tongue vertex positions, and / or a set of PCA coefficients associated with a mesh from the additional feature set. The output may additionally be converted by a renderer into animation frames.

[0115] When the detection engine 122 determines that the speech rate exceeds one or more thresholds, the processing engine 124 performs operation 514, in which the processing engine 124 upsamples the features generated in operation 504 based on the exceeded thresholds to generate an upsampled feature set. For example, the processing engine 124 may receive the features generated in operation 504, the highest threshold exceeded by the features, and / or the upsampling factor associated with the highest threshold from the detection engine 122. The processing engine 124 may also perform bilinear upsampling on the features by the upsampling factor to generate a larger upsampled feature set.

[0116] In operation 516, the processing engine 124 extracts a subset of the upsampled features. For example, the processing engine 124 may extract a certain number of features from the temporal center of the upsampled features (e.g., the same number of features as the feature set generated in operation 504). Operation 510 is then performed by the pronunciation network 210 using the extracted subset of the upsampled features to generate an additional feature set, and operation 512 is performed by the output network 212 to convert the additional feature set into a corresponding facial animation output.

[0117] After generating the facial animation output for the audio clip, the system 200 performs operation 518, wherein the system 200 determines whether to continue generating the facial animation output. For example, the system 200 may determine that additional facial animation output is to be generated while the additional audio clip remains extracted from the audio sample. When the system 200 determines that additional facial animation output is to be generated, the system 200 repeats operations 502, 504, 506, 508, 510, 512, 514, and / or 516 for each audio clip. For example, the system 200 may iteratively perform operation 502 in a manner that results in extracting a sequence of audio clips of the same length and a certain stride and / or offset within the audio sample. The system 200 may also perform operations 504 and 506 for each audio clip to determine the individual speech rate of the audio clip, the aggregate speech rate of a series of audio clips including the audio clip, and the feature set of the audio clip. The system 200 may then perform operation 508 to determine whether the speech rate associated with each audio clip exceeds any threshold. If the speech rate does not exceed any threshold, the system 200 may perform operations 510 and 512 using the features generated in operation 504. If the speech rate exceeds one or more thresholds, the system 200 may perform operations 514 and 516 to convert the features generated in operation 504 into upsampled features and cropped features, and then use the upsampled features and cropped features to perform operations 510 and 512. Thus, the system may continue to generate facial animation output until the entire audio sample has been processed.

[0118] It should be understood that this and other arrangements described herein are only described as examples. In addition to or in place of the arrangements and elements shown, other arrangements and elements (e.g., machines, interfaces, functions, orders, functional groupings, etc.) may be used, and some elements may be completely omitted. Further, many elements described herein are functional entities that may be implemented as discrete or distributed components or in combination with other components, and in any suitable combination and position. Various functions described herein as being performed by an entity may be performed by hardware, firmware, and / or software. For example, various functions may be implemented by a processor executing instructions stored in a memory. As discussed, aspects of the various methods presented herein may be sufficiently lightweight to be performed in real time or near real time on devices such as client devices (such as personal computers or game consoles). Such processing may be performed on content (e.g., streaming sensor data or other content received through at least one network) generated on the client device or received by it or received from an external source. In some instances, the processing and / or determination of this content may be performed by one of these other devices, systems, or entities, and then provided to the client device (or another such receiver) for presentation or another such use.

[0119] As an example, Figure 6An example network configuration 600 is shown that can be used to provide, generate, modify, encode and / or transmit data or other such content. In at least one embodiment, a client device 602 can generate or receive data for a session using components of a content application 604 on the client device 602, and the data is stored locally on the client device. In at least one embodiment, a content application 624 executed on a server 620 (e.g., a cloud server or edge server) can initiate a session associated with at least the client device 602, such as using a session manager and user data stored in a user database, and can cause content to be determined by a content manager 626. The content manager 626 can work with an audio to emotion module 628 or system to determine motion corresponding to input audio taking into account emotion or style data, and with a renderer 630 to generate images, videos, or other visual presentation data using assets (e.g., character meshes) from an asset database 632, to the extent permissible as determined by a rights manager 634 or other such component or service. At least a portion of the generated content (separate and distinct from the asset itself) may be transmitted to the client device 602 using an appropriate transport manager 622 for transmission via download, streaming, or another such transmission channel. An encoder may be used to encode and / or compress at least some of the data before transmission to the client device 602. In at least one embodiment, a client device 602 receiving such content may provide the content to a corresponding content application 604, which may also or alternatively include a graphical user interface 610, an audio to emotion component 612, and a renderer 614 for generating audio-driven animations or images. A decoder may also be used to decode data received over a network 640 for presentation via the client device 602, such as presentation of images or video content via a display 606 and presentation of audio (such as sounds and music) via at least one audio playback device 608 (such as a speaker or headphones). In at least one embodiment, at least some of the content may already be stored on, rendered on, or accessible to the client device 602, such that no transmission over the network 640 is required for at least that portion of the content, such as the content may have been previously downloaded or stored locally on a hard drive or optical disk. In at least one embodiment, a transmission mechanism such as data streaming may be used to transmit the content from the server 620 to the client device 602. In at least one embodiment, at least a portion of the content may be obtained or streamed from another source such as a third party service 660 or other client device 650, which may also include a content application 662 for generating or providing the content.In at least one embodiment, portions of this functionality may be performed using multiple computing devices or multiple processors within one or more computing devices (such as may include a combination of CPUs and GPUs).

[0120] In this example, the client devices may include any appropriate computing devices, such as desktop computers, laptop computers, set-top boxes, streaming devices, game consoles, smart phones, tablet computers, VR / AR / MR head displays, VR / AR / MR goggles, wearable computers, or smart TVs. Each client device may submit a request through at least one wired or wireless network, such as the Internet, Ethernet, local area network (LAN), or cellular network and other such options. In this example, the requests may be submitted to an address associated with a cloud provider, and the cloud provider may operate or control one or more electronic resources in the cloud provider environment, such as a data center or server farm. In at least one embodiment, the request may be received or processed by at least one edge server located on the edge of the network and outside of at least one security layer associated with the cloud provider environment. In this way, by enabling client devices to interact with closer servers, latency can be reduced while also improving the security of resources in the cloud provider environment.

[0121] In at least one embodiment, such a system can be used to perform graphics rendering operations. In other embodiments, such a system can be used for other purposes, such as for providing image or video content to test or verify autonomous machine applications, or for performing deep learning operations. In at least one embodiment, such a system can be implemented using an edge device, or can include one or more virtual machines (VMs). In at least one embodiment, such a system can be implemented at least in part in a data center or at least in part using cloud computing resources.

[0122] Reasoning and training logic

[0123] Fig. 7A Inference and / or training logic 715 is shown for performing reasoning and / or training operations associated with one or more embodiments. Fig. 7A and / or Figure 7B Details regarding the inference and / or training logic 715 are provided.

[0124] In at least one embodiment, the inference and / or training logic 715 may include, but is not limited to, code and / or data storage 701 for storing forward and / or output weights and / or input / output data, and / or other parameters for configuring neurons or layers of a neural network trained and / or used for inference in aspects of one or more embodiments. In at least one embodiment, the training logic 715 may include or be coupled to code and / or data storage 701 for storing graph code or other software to control timing and / or sequence, where weights and / or other parameter information is loaded to configure logic, including integer and / or floating point units (collectively referred to as arithmetic logic units (ALUs)). In at least one embodiment, the code (such as graph code) loads weights or other parameter information into a processor ALU based on the architecture of the neural network to which the code corresponds. In at least one embodiment, the code and / or data storage 701 stores input / output data during training and / or inference using aspects of one or more embodiments and / or weight parameters during forward propagation of each layer of a neural network trained or used in conjunction with one or more embodiments. In at least one embodiment, any portion of code and / or data storage 701 may be included within other on-chip or off-chip data storage, including a processor's L1, L2, or L3 cache or system memory.

[0125] In at least one embodiment, any portion of code and / or data storage 701 may be internal or external to one or more processors or other hardware logic devices or circuits. In at least one embodiment, code and / or data storage 701 may be cache memory, dynamic random access memory ("DRAM"), static random access memory ("SRAM"), non-volatile memory (e.g., flash memory), or other storage. In at least one embodiment, the choice of whether code and / or data storage 701 is internal or external to a processor, for example, or consists of DRAM, SRAM, flash memory, or some other storage type, may depend on the available storage space on-chip or off-chip, the latency requirements of the training and / or inference functions being performed, the batch size of the data used in inference and / or training of the neural network, or some combination of these factors.

[0126] In at least one embodiment, the inference and / or training logic 715 may include, but is not limited to, code and / or data storage 705 for storing reverse and / or output weights and / or input / output data corresponding to neurons or layers of a neural network trained and / or used for inference in aspects of one or more embodiments. In at least one embodiment, during training and / or inference using aspects of one or more embodiments, the code and / or data storage 705 stores weight parameters and / or input / output data for each layer of a neural network trained or used in conjunction with one or more embodiments during back propagation of input / output data and / or weight parameters. In at least one embodiment, the training logic 715 may include or be coupled to code and / or data storage 705 for storing graph code or other software to control timing and / or sequence, wherein weights and / or other parameter information is loaded to configure logic including integer and / or floating point units (collectively referred to as arithmetic logic units (ALUs)). In at least one embodiment, code (such as graph code) loads weights or other parameter information into a processor ALU based on the architecture of the neural network to which the code corresponds. In at least one embodiment, any portion of code and / or data storage 705 may be included with other on-chip or off-chip data storage, including the processor's L1, L2, or L3 cache or system memory. In at least one embodiment, any portion of code and / or data storage 705 may be internal or external to one or more processors or other hardware logic devices or circuits. In at least one embodiment, code and / or data storage 705 may be cache memory, DRAM, SRAM, non-volatile memory (e.g., flash memory), or other storage. In at least one embodiment, the choice of whether code and / or data storage 705 is internal or external to the processor, for example, consisting of DRAM, SRAM, flash memory, or some other storage type, depends on whether the available storage is on-chip or off-chip, the latency requirements of the training and / or inference functions being performed, the data batch size used in the inference and / or training of the neural network, or some combination of these factors.

[0127] In at least one embodiment, code and / or data storage 701 and code and / or data storage 705 may be separate storage structures. In at least one embodiment, code and / or data storage 701 and code and / or data storage 705 may be the same storage structure. In at least one embodiment, code and / or data storage 701 and code and / or data storage 705 may be partially the same storage structure and partially separate storage structures. In at least one embodiment, any portion of code and / or data storage 701 and code and / or data storage 705 may be included with other on-chip or off-chip data storage, including the processor's L1, L2, or L3 cache or system memory.

[0128] In at least one embodiment, the inference and / or training logic 715 may include, but is not limited to, one or more arithmetic logic units (“ALUs”) 710 (including integer and / or floating point units) for performing logical and / or mathematical operations based at least in part on or as directed by training and / or inference code (e.g., graph code), the results of which may produce activations (e.g., output values ​​from a layer or neuron within a neural network) stored in activation storage 720, which is a function of input / output and / or weight parameter data stored in code and / or data storage 701 and / or code and / or data storage 705. In at least one embodiment, activations are performed in response to executing instructions or other code, linear algebra and / or matrix-based mathematics performed by ALU 710 to generate activations stored in activation storage 720, where weight values ​​stored in code and / or data storage 705 and / or in code and / or data storage 701 are used as operands with other values, such as bias values, gradient information, momentum values, or other parameters or hyperparameters, any or all of which may be stored in code and / or data storage 705 or code and / or data storage 701 or other on-chip or off-chip storage.

[0129] In at least one embodiment, one or more ALUs 710 are included in one or more processors or other hardware logic devices or circuits, while in another embodiment, one or more ALUs 710 may be outside a processor or other hardware logic device or circuit that uses them (e.g., a coprocessor). In at least one embodiment, one or more ALUs 710 may be included within an execution unit of a processor, or otherwise included in a group of ALUs accessible by an execution unit of a processor, which may be within the same processor or distributed between different processors of different types (e.g., a central processing unit, a graphics processing unit, a fixed function unit, etc.). In at least one embodiment, code and / or data storage 701, code and / or data storage 705, and activation storage 720 may be on the same processor or other hardware logic device or circuit, while in another embodiment, they may be in different processors or other hardware logic devices or circuits or some combination of the same and different processors or other hardware logic devices or circuits. In at least one embodiment, any portion of activation storage 720 may be included with other on-chip or off-chip data storage, including the processor's L1, L2, or L3 cache or system memory. Additionally, inference and / or training code may be stored with other code accessible to a processor or other hardware logic or circuitry and may be retrieved and / or processed using the processor's fetch, decode, schedule, execute, exit, and / or other logic circuitry.

[0130] In at least one embodiment, activation storage 720 may be cache memory, DRAM, SRAM, non-volatile memory (e.g., flash memory), or other storage. In at least one embodiment, activation storage 720 may be completely or partially internal or external to one or more processors or other logic circuits. In at least one embodiment, activation storage 720 may be selected to be internal or external to a processor, for example, or include DRAM, SRAM, flash memory, or other storage types, depending on the storage available on-chip or off-chip, the latency requirements for performing training and / or inference functions, the batch size of data used in inferencing and / or training neural networks, or some combination of these factors. In at least one embodiment, Fig. 7A The inference and / or training logic 715 shown in FIG. 7 may be used in conjunction with an application specific integrated circuit (“ASIC”), such as the ASIC from Google. Processing unit from Graphcore TM Inference Processing Unit (IPU) from Intel Corp (e.g., "LakeCrest") processor. In at least one embodiment, Fig. 7A The illustrated inference and / or training logic 715 may be used in conjunction with central processing unit (“CPU”) hardware, graphics processing unit (“GPU”) hardware, or other hardware such as a field programmable gate array (“FPGA”).

[0131] Figure 7B Inference and / or training logic 715 is shown in accordance with at least one or more embodiments. In at least one embodiment, the reasoning and / or training logic 715 may include, but is not limited to, hardware logic in which computing resources are dedicated or otherwise uniquely used in conjunction with weight values ​​or other information corresponding to one or more layers of neurons within a neural network. In at least one embodiment, the reasoning and / or training logic 715 includes, but is not limited to, code and / or data storage 701 and code and / or data storage 705, which may be used to store code (e.g., graph code), weight values, and / or other information, including bias values, gradient information, momentum values, and / or other parameter or hyperparameter information. Figure 7B In at least one embodiment shown in , each of code and / or data storage 701 and code and / or data storage 705 is associated with a dedicated computing resource (e.g., computing hardware 702 and computing hardware 706), respectively. In at least one embodiment, each of computing hardware 702 and computing hardware 706 includes one or more ALUs that perform mathematical functions (e.g., linear algebraic functions) solely on the information stored in code and / or data storage 701 and code and / or data storage 705, respectively, and the results of performing the functions are stored in activation storage 720.

[0132] In at least one embodiment, each of the code and / or data stores 701 and 705 and the corresponding computing hardware 702 and 706 corresponds to a different layer of the neural network, such that activations from one "storage / compute pair 701 / 702" of the code and / or data store 701 and computing hardware 702 are provided as inputs to the next "storage / compute pair 705 / 706" of the code and / or data store 705 and computing hardware 706, so as to reflect the conceptual organization of the neural network. In at least one embodiment, each storage / compute pair 701 / 702 and 705 / 706 can correspond to more than one neural network layer. In at least one embodiment, additional storage / compute pairs (not shown) can be included in the inference and / or training logic 715 after or in parallel with the storage / compute pairs 701 / 702 and 705 / 706.

[0133] Data Center

[0134] Figure 8 An example data center 800 is shown in which at least one embodiment may be used. In at least one embodiment, the data center 800 includes a data center infrastructure layer 810, a framework layer 820, a software layer 830, and an application layer 840.

[0135] In at least one embodiment, Figure 8 As shown, the data center infrastructure layer 810 may include a resource coordinator 812, grouped computing resources 814, and node computing resources ("node CRs") 816(1)-816(N), where "N" represents any positive integer. In at least one embodiment, the node CRs 816(1)-816(N) may include, but are not limited to, any number of central processing units ("CPUs") or other processors (including accelerators, field programmable gate arrays (FPGAs), graphics processors, etc.), memory devices (e.g., dynamic read-only memories), storage devices (e.g., solid-state drives or disk drives), network input / output ("NWI / O") devices, network switches, virtual machines ("VMs"), power modules and cooling modules, etc. In at least one embodiment, one or more of the node CRs 816(1)-816(N) may be a server having one or more of the above computing resources.

[0136] In at least one embodiment, the grouped computing resources 814 may include a separate grouping (not shown) of node CRs housed in one or more racks, or many racks (also not shown) housed in data centers at various geographic locations. The separate grouping of node CRs within the grouped computing resources 814 may include computing, networks, memory, or storage resources that can be configured or allocated to support groupings of one or more workloads. In at least one embodiment, several node CRs including a CPU or processor may be grouped in one or more racks to provide computing resources to support one or more workloads. In at least one embodiment, one or more racks may also include any number of power modules, cooling modules, and network switches in any combination.

[0137] In at least one embodiment, resource coordinator 812 may configure or otherwise control one or more node CRs 816(1)-816(N) and / or grouped computing resources 814. In at least one embodiment, resource coordinator 812 may include a software design infrastructure ("SDI") management entity for data center 800. In at least one embodiment, resource coordinator 108 may include hardware, software, or some combination thereof.

[0138] In at least one embodiment, Figure 8 As shown, the framework layer 820 includes a job scheduler 822, a configuration manager 824, a resource manager 826, and a distributed file system 828. In at least one embodiment, the framework layer 820 may include a framework that supports software 832 of the software layer 830 and / or one or more applications 842 of the application layer 840. In at least one embodiment, the software 832 or the application 842 may include web-based service software or applications, such as those provided by Amazon Web Services, Google Cloud, and Microsoft Azure. In at least one embodiment, the framework layer 820 may be, but is not limited to, a free and open source software web application framework, such as Apache Spark, which can utilize the distributed file system 828 for large-scale data processing (e.g., "big data"). TM(hereinafter referred to as "Spark"). In at least one embodiment, the job scheduler 832 may include a Spark driver to facilitate scheduling of workloads supported by various layers of the data center 800. In at least one embodiment, the configuration manager 824 may be able to configure different layers, such as the software layer 830 and the framework layer 820 including Spark and a distributed file system 828 for supporting large-scale data processing. In at least one embodiment, the resource manager 826 can manage cluster or group computing resources mapped to or allocated to support the distributed file system 828 and the job scheduler 822. In at least one embodiment, the cluster or group computing resources may include group computing resources 814 on the data center infrastructure layer 810. In at least one embodiment, the resource manager 826 can coordinate with the resource coordinator 812 to manage these mapped or allocated computing resources.

[0139] In at least one embodiment, software 832 included in software layer 830 may be used by at least a portion of node CRs 816(1)-816(N), grouped computing resources 814, and / or software used by distributed file system 828 of framework layer 820. One or more types of software 832 may include, but are not limited to, Internet web page search software, email virus scanning software, database software, and streaming video content software.

[0140] In at least one embodiment, one or more applications 842 included in the application layer 840 may include one or more types of applications used by at least a portion of the node CRs 816(1)-816(N), the grouped computing resources 814, and / or the distributed file system 828 of the framework layer 820. The one or more types of applications may include, but are not limited to, any number of genomics applications, cognitive computing, and machine learning applications, including training or inference software, machine learning framework software (e.g., PyTorch, TensorFlow, Caffe, etc.), or other machine learning applications used in conjunction with one or more embodiments.

[0141] In at least one embodiment, any of the configuration manager 824, resource manager 826, and resource coordinator 812 can implement any number and type of self-modification actions based on any number and type of data acquired in any technically feasible manner. In at least one embodiment, the self-modification actions can relieve a data center operator of the data center 800 from making potentially bad configuration decisions and can avoid underutilized and / or poorly performing portions of the data center.

[0142] In at least one embodiment, the data center 800 may include tools, services, software, or other resources to train one or more machine learning models or use one or more machine learning models to predict or infer information according to one or more embodiments described herein. For example, in at least one embodiment, a machine learning model may be trained by calculating weight parameters according to a neural network architecture using the software and computing resources described above with respect to the data center 800. In at least one embodiment, by using weight parameters calculated by one or more training techniques described herein, information may be inferred or predicted using trained machine learning models corresponding to one or more neural networks using the resources described above with respect to the data center 800.

[0143] In at least one embodiment, the data center 800 can use a CPU, an application-specific integrated circuit (ASIC), a GPU, an FPGA, or other hardware to use the above resources to perform training and / or reasoning. In addition, one or more of the above software and / or hardware resources can be configured as a service to allow users to train or perform information reasoning, such as image recognition, speech recognition, or other artificial intelligence services.

[0144] In at least one embodiment, the reasoning and / or training logic 715 may be Figure 8 The system is used to infer or predict operations based at least in part on weight parameters calculated using neural network training operations, neural network functions and / or architectures, or neural network use cases described herein. Such components can be used to generate facial animations from audio data, including accurate depictions of emotional states.

[0145] Computer Systems

[0146] Fig. 9 900 is a block diagram illustrating an exemplary computer system according to various embodiments, which may be a system of interconnected devices and components, a system on a chip (SOC), or some combination thereof formed with a processor, which may include an execution unit to execute instructions. In at least one embodiment, in accordance with the present disclosure, such as the embodiments described herein, the computer system 900 may include, but is not limited to, components, such as a processor 902, whose execution unit includes logic to execute algorithms for process data. In at least one embodiment, the computer system 900 may include a processor, such as a processor available from Intel Corporation of Santa Clara, California. Processor family, Xeon TM , XScale TM and / or StrongARMTM , Core TM or Nervana TM microprocessor, although other systems (including PCs with other microprocessors, engineering workstations, set-top boxes, etc.) may also be used. In at least one embodiment, computer system 900 may execute a version of the WINDOWS operating system available from Microsoft Corporation of Redmond, Wash., although other operating systems (e.g., UNIX and Linux), embedded software, and / or graphical user interfaces may also be used.

[0147] Embodiments may be used in other devices, such as handheld devices and embedded applications. Some examples of handheld devices include cellular phones, Internet Protocol (Internet Protocol) devices, digital cameras, personal digital assistants ("PDAs"), and handheld PCs. In at least one embodiment, the embedded application may include a microcontroller, a digital signal processor ("DSP"), a system on a chip, a network computer ("NetPC"), a set-top box, a network hub, a wide area network ("WAN") switch, or any other system that can execute one or more instructions according to at least one embodiment.

[0148] In at least one embodiment, the computer system 900 may include, but is not limited to, a processor 902, which may include, but is not limited to, one or more execution units 908 to perform machine learning model training and / or reasoning according to the techniques described herein. In at least one embodiment, the computer system 900 is a single-processor desktop or server system, but in another embodiment, the computer system 900 may be a multi-processor system. In at least one embodiment, the processor 902 may include, but is not limited to, a complex instruction set computer ("CISC") microprocessor, a reduced instruction set computing ("RISC") microprocessor, a very long instruction word ("VLIW") microprocessor, a processor that implements an instruction set combination, or any other processor device, such as a digital signal processor. In at least one embodiment, the processor 902 may be coupled to a processor bus 910, which may transmit data signals between the processor 902 and other components in the computer system 900.

[0149] In at least one embodiment, processor 902 may include, but is not limited to, a level 1 ("L1") internal cache memory ("cache") 904. In at least one embodiment, processor 902 may have a single internal cache or multiple levels of internal cache. In at least one embodiment, the cache memory may reside external to processor 902. Other embodiments may also include a combination of internal and external caches, depending on the particular implementation and needs. In at least one embodiment, register file 906 may store different types of data in various registers, including, but not limited to, integer registers, floating point registers, status registers, and instruction pointer registers.

[0150] In at least one embodiment, logic execution unit 908 that performs integer and floating point operations is also located in processor 902, including, but not limited to. In at least one embodiment, processor 902 may also include a microcode ("ucode") read-only memory ("ROM") for storing microcode for certain macroinstructions. In at least one embodiment, execution unit 908 may include logic for processing a packed instruction set 909. In at least one embodiment, by including a packed instruction set 909 in the instruction set of a general purpose processor, and associated circuitry to execute the instructions, operations used by many multimedia applications may be performed using packed data in processor 902. In one or more embodiments, many multimedia applications may be executed faster and more efficiently by using the full width of the processor's data bus to perform operations on packed data, which may not require the transfer of smaller units of data on the processor's data bus to perform one or more operations one data element at a time.

[0151] In at least one embodiment, execution unit 908 may also be used in microcontrollers, embedded processors, graphics devices, DSPs, and other types of logic circuits. In at least one embodiment, computer system 900 may include, but is not limited to, memory 920. In at least one embodiment, memory 920 may be implemented as a dynamic random access memory ("DRAM") device, a static random access memory ("SRAM") device, a flash memory device, or other storage device. In at least one embodiment, memory 920 may store instructions 919 and / or data 921 represented by data signals that may be executed by processor 902.

[0152] In at least one embodiment, the system logic chip can be coupled to the processor bus 910 and the memory 920. In at least one embodiment, the system logic chip can include, but is not limited to, a memory controller hub ("MCH") 916, and the processor 902 can communicate with the MCH 916 via the processor bus 910. In at least one embodiment, the MCH 916 can provide a high-bandwidth memory path 918 to the memory 920 for instruction and data storage and for storage of graphics commands, data, and textures. In at least one embodiment, the MCH 916 can initiate data signals between the processor 902, the memory 920, and other components in the computer system 900, and bridge data signals between the processor bus 910, the memory 920, and the system I / O 922. In at least one embodiment, the system logic chip can provide a graphics port for coupling to a graphics controller. In at least one embodiment, the MCH 916 can be coupled to the memory 920 via a high-bandwidth memory path 918, and the graphics / video card 912 can be coupled to the MCH 916 via an Accelerated Graphics Port ("AGP") interconnect 914.

[0153] In at least one embodiment, the computer system 900 may use a system I / O 922, which is a proprietary hub interface bus to couple the MCH 916 to an I / O controller hub ("ICH") 930. In at least one embodiment, the ICH 930 may provide direct connections to certain I / O devices through a local I / O bus. In at least one embodiment, the local I / O bus may include, but is not limited to, a high-speed I / O bus used to connect peripheral devices to the memory 920, chipset, and processor 902. Examples may include, but are not limited to, an audio controller 929, a firmware hub ("FlashBIOS") 928, a wireless transceiver 926, a data store 924, a traditional I / O controller 923 including a user input and keyboard interface, a serial expansion port 927 (e.g., a universal serial bus (USB) port), and a network controller 934. The data store 924 may include a hard drive, a floppy drive, a CD-ROM device, a flash memory device, or other mass storage device.

[0154] In at least one embodiment, Fig. 9 The system is shown as comprising interconnected hardware devices or "chips", while in other embodiments, Fig. 9An exemplary system on chip (SoC) may be shown. In at least one embodiment, the devices may be interconnected with a proprietary interconnect, a normalized interconnect (e.g., PCIe), or some combination thereof. In at least one embodiment, one or more components of computer system 900 are interconnected using a compute express link (CXL) interconnect.

[0155] In at least one embodiment, the reasoning and / or training logic 715 may be Fig. 9 The system is used to infer or predict operations based at least in part on weight parameters calculated using neural network training operations, neural network functions and / or architectures, or neural network use cases described herein. Such components can be used to generate facial animations from audio data, including accurate depictions of emotional states.

[0156] Fig.10 1 is a block diagram illustrating an electronic device 1000 for utilizing a processor 1010 according to various embodiments. In at least one embodiment, the electronic device 1000 may be, for example but not limited to, a notebook computer, a tower server, a rack server, a blade server, a laptop computer, a desktop computer, a tablet computer, a mobile device, a phone, an embedded computer, or any other suitable electronic device.

[0157] In at least one embodiment, the electronic device 1000 may include, but is not limited to, a processor 1010 communicatively coupled to any suitable number or kind of components, peripherals, modules, or devices. In at least one embodiment, the processor 1010 is coupled using a bus or interface, such as an Ic bus, a system management bus ("SMBus"), a low pin count (LPC) bus, a serial peripheral interface ("SPI"), a high-definition audio ("HDA") bus, a serial advanced technology attachment ("SATA") bus, a universal serial bus ("USB") (versions 1, 2, 3), or a universal asynchronous receiver / transmitter ("UART") bus.

[0158] In at least one embodiment, Fig.10 A system is shown that includes interconnected hardware devices or "chips", while in other embodiments, Fig.10 An exemplary system on chip (SoC) may be shown. In at least one embodiment, Fig.10 The devices shown in can be interconnected with a proprietary interconnect, a normalized interconnect (e.g., PCIe), or some combination thereof. In at least one embodiment, Fig.10 One or more components of the system are interconnected using Compute Express Link (CXL) interconnect lines.

[0159] In at least one embodiment, Fig.10It may include a display 1024, a touch screen 1025, a touch pad 1030, a near field communication unit (“NFC”) 1045, a sensor hub 1040, a thermal sensor 1046, a fast chipset (“EC”) 1035, a trusted platform module (“TPM”) 1038, a BIOS / firmware / flash memory (“BIOS, FWFlash”) 1022, a DSP 1060, a drive 1020 (e.g., a solid state disk (“SSD”) or a hard disk drive (“HDD”)), a wireless local area network unit (“WLAN”) 1050, a Bluetooth unit 1052, a wireless wide area network unit (“WWAN”) 1056, a global positioning system (GPS) 1055, a camera (“USB 3.0 camera”) 1054 (e.g., a USB 3.0 camera), and / or a low power double data rate (“LPDDR”) memory unit (“LPDDR3”) 1015 implemented with, for example, the LPDDR3 standard. These components may each be implemented in any suitable manner.

[0160] In at least one embodiment, other components may be communicatively coupled to the processor 1010 through the components described above. In at least one embodiment, the accelerometer 1041, ambient light sensor (“ALS”) 1042, compass 1043, and gyroscope 1044 may be communicatively coupled to the sensor hub 1040. In at least one embodiment, the thermal sensor 1039, fan 1037, keyboard 1036, and touchpad 1030 may be communicatively coupled to the EC 1035. In at least one embodiment, the speaker 1063, earphone 1064, and microphone (“mic”) 1065 may be communicatively coupled to the audio unit (“audio codec and class D amplifier”) 1062, which in turn may be communicatively coupled to the DSP 1060. In at least one embodiment, the audio unit 1062 may include, for example, but not limited to, an audio encoder / decoder (“codec”) and a class D amplifier. In at least one embodiment, the SIM card (“SIM”) 1057 may be communicatively coupled to the WWAN unit 1056. In at least one embodiment, components such as the WLAN unit 1050 and the Bluetooth unit 1052 and the WWAN unit 1056 may be implemented as a next generation form factor (NGFF).

[0161] In at least one embodiment, the reasoning and / or training logic 715 may be Fig.10 A system for reasoning or predicting operations based at least in part on weight parameters computed using neural network training operations, neural network functions and / or architectures, or neural network use cases described herein. Such components can be used to generate facial animations from audio data, including accurate depictions of emotional states.

[0162] Fig.111100 is a block diagram of a processing system according to various embodiments. In at least one embodiment, system 1100 includes one or more processors 1102 and one or more graphics processors 1108, and can be a single-processor desktop system, a multi-processor workstation system, or a server system with a large number of processors 1102 or processor cores 1107. In at least one embodiment, system 1100 is a processing platform incorporated within a system-on-chip (SoC) integrated circuit for use in a mobile, handheld, or embedded device.

[0163] In at least one embodiment, the system 1100 may include or be incorporated into a server-based gaming platform, including a gaming console, a mobile gaming console, a handheld gaming console, or an online gaming console for gaming and media consoles. In at least one embodiment, the system 1100 is a mobile phone, a smart phone, a tablet computing device, or a mobile Internet device. In at least one embodiment, the processing system 1100 may also include a wearable device coupled to or integrated in a wearable device, such as a smart watch wearable device, a smart glasses device, an augmented reality device, or a virtual reality device. In at least one embodiment, the processing system 1100 is a television or set-top box device having one or more processors 1102 and a graphical interface generated by one or more graphics processors 1108.

[0164] In at least one embodiment, one or more processors 1102 each include one or more processor cores 1107 to process instructions that, when executed, perform operations for system and user software. In at least one embodiment, each of the one or more processor cores 1107 is configured to process a specific instruction group 1109. In at least one embodiment, the instruction group 1109 can facilitate complex instruction set computing (CISC), reduced instruction set computing (RISC), or calculate by very long instruction words (VLIW). In at least one embodiment, the processor cores 1107 can each process different instruction groups 1109, which can include instructions that help emulate other instruction groups. In at least one embodiment, the processor cores 1107 can also include other processing devices, such as digital signal processors (DSPs).

[0165] In at least one embodiment, the processor 1102 includes a cache memory 1104. In at least one embodiment, the processor 1102 may have a single internal cache or multiple levels of internal cache. In at least one embodiment, the cache memory is shared between various components of the processor 1102. In at least one embodiment, the processor 1102 also uses an external cache (e.g., a level 3 (L3) cache or a last level cache (LLC)) (not shown), which can be shared between the processor cores 1107 using known cache coherence techniques. In at least one embodiment, the processor 1102 additionally includes a register file 1106, and the processor may include different types of registers (e.g., integer registers, floating point registers, status registers, and instruction pointer registers) for storing different types of data. In at least one embodiment, the register file 1106 may include general registers or other registers.

[0166] In at least one embodiment, one or more processors 1102 are coupled to one or more interface buses 1110 to transmit communication signals, such as address, data, or control signals, between the processor 1102 and other components in the system 1100. In at least one embodiment, the interface bus 1110 may be a processor bus in one embodiment, such as a version of a direct media interface (DMI) bus. In at least one embodiment, the interface bus 1110 is not limited to a DMI bus, and may include one or more peripheral component interconnect buses (e.g., PCI, PCI Express), a memory bus, or other types of interface buses. In at least one embodiment, the processor 1102 includes an integrated memory controller 1116 and a platform controller hub 1130. In at least one embodiment, the memory controller 1116 facilitates communication between memory devices and other components of the processing system 1100, while the platform controller hub (PCH) 1130 provides connections to I / O devices through a local I / O bus.

[0167] In at least one embodiment, the memory device 1120 may be a dynamic random access memory (DRAM) device, a static random access memory (SRAM) device, a flash memory device, a phase change memory device, or have appropriate performance to be used as a processor memory. In at least one embodiment, the memory device 1120 may be used as a system memory of the processing system 1100 to store data 1122 and instructions 1121 for use when one or more processors 1102 execute applications or processes. In at least one embodiment, the memory controller 1116 is also coupled to an optional external graphics processor 1112, which may communicate with one or more graphics processors 1108 in the processor 1102 to perform graphics and media operations. In at least one embodiment, a display device 1111 may be connected to the processor 1102. In at least one embodiment, the display device 1111 may include one or more of the internal display devices, such as in a mobile electronic device or laptop device or an external display device connected via a display interface (e.g., DisplayPort, etc.). In at least one embodiment, the display device 1111 may include a head mounted display (HMD), such as a stereoscopic display device used in virtual reality (VR) applications or augmented reality (AR) applications.

[0168] In at least one embodiment, the platform controller hub 1130 enables peripheral devices to be connected to the storage device 1120 and the processor 1102 via a high-speed I / O bus. In at least one embodiment, the I / O peripherals include, but are not limited to, an audio controller 1146, a network controller 1134, a firmware interface 1128, a wireless transceiver 1126, a touch sensor 1125, a data storage device 1124 (e.g., a hard drive, flash memory, etc.). In at least one embodiment, the data storage device 1124 can be connected via a storage interface (e.g., SATA) or via a peripheral bus, such as a peripheral component interconnect bus (e.g., PCI, PCIe). In at least one embodiment, the touch sensor 1125 can include a touch screen sensor, a pressure sensor, or a fingerprint sensor. In at least one embodiment, the wireless transceiver 1126 can be a Wi-Fi transceiver, a Bluetooth transceiver, or a mobile network transceiver, such as a 3G, 4G, or long-term evolution (LTE) transceiver. In at least one embodiment, the firmware interface 1128 enables communication with the system firmware and can be, for example, a unified extensible firmware interface (UEFI). In at least one embodiment, the network controller 1134 can enable a network connection to a wired network. In at least one embodiment, a high-performance network controller (not shown) is coupled to the interface bus 1110. In at least one embodiment, the audio controller 1146 is a multi-channel high-definition audio controller. In at least one embodiment, the processing system 1100 includes an optional legacy I / O controller 1140 for coupling legacy (e.g., Personal System 2 (PS / 2)) devices to the system 1100. In at least one embodiment, the platform controller hub 1130 can also be connected to one or more universal serial bus (USB) controllers 1142, which connect input devices such as a keyboard and mouse 1143 combination, a camera 1144, or other USB input devices.

[0169] In at least one embodiment, instances of memory controller 1116 and platform controller hub 1130 may be integrated into a discrete external graphics processor, such as external graphics processor 1112. In at least one embodiment, platform controller hub 1130 and / or memory controller 1116 may be external to one or more processors 1102. For example, in at least one embodiment, system 1100 may include external memory controller 1116 and platform controller hub 1130, which may be configured as a memory controller hub and a peripheral controller hub in a system chipset that communicates with processor 1102.

[0170] In at least one embodiment, some or all of the inference and / or training logic 715 may be incorporated into the graphics processor 1100. For example, in at least one embodiment, the training and / or inference techniques described herein may use one or more ALUs that are embodied in the graphics processor. In addition, in at least one embodiment, the inference and / or training operations described herein may use a CPU other than the ALU. Fig. 7A or Figure 7B In at least one embodiment, the weight parameters may be stored in on-chip or off-chip memory and / or registers (shown or not shown), which configure the ALU of the graphics processor to execute one or more machine learning algorithms, neural network architectures, use cases, or training techniques described herein. Such components may be used to generate facial animations from audio data, including accurate depictions of emotional states.

[0171] Fig.12 is a block diagram of a processor 1200 having one or more processor cores 1202A-1202N, an integrated memory controller 1214, and an integrated graphics processor 1208 in accordance with various embodiments. In at least one embodiment, the processor 1200 may include additional cores, up to and including the additional core 1202N represented by the dashed box. In at least one embodiment, each processor core 1202A-1202N includes one or more internal cache units 1204A-1204N. In at least one embodiment, each processor core may also have access to one or more shared cache units 1206.

[0172] In at least one embodiment, the internal cache units 1204A-1204N and the shared cache unit 1206 represent a cache memory hierarchy within the processor 1200. In at least one embodiment, the cache memory units 1204A-1204N may include at least one level of instruction and data cache within each processor core and one or more levels of cache in a shared mid-level cache, such as level 2 (L2), level 3 (L3), level 4 (L4), or other levels of cache, where the highest level of cache before external memory is categorized as LLC. In at least one embodiment, cache coherence logic maintains coherence between the various cache units 1206 and 1204A-1204N.

[0173] In at least one embodiment, the processor 1200 may also include a set of one or more bus controller units 1216 and a system agent core 1210. In at least one embodiment, the one or more bus controller units 1216 manage a set of peripheral buses, such as one or more PCI or PCIe buses. In at least one embodiment, the system agent core 1210 provides management functions for various processor components. In at least one embodiment, the system agent core 1210 includes one or more integrated memory controllers 1214 to manage access to various external memory devices (not shown).

[0174] In at least one embodiment, one or more processor cores 1202A-1202N include support for multiple threads simultaneously. In at least one embodiment, system agent core 1210 includes components for coordinating and operating cores 1202A-1202N during multithreaded processing. In at least one embodiment, system agent core 1210 may additionally include a power control unit (PCU) that includes logic and components for regulating one or more power states of processor cores 1202A-1202N and graphics processor 1208.

[0175] In at least one embodiment, the processor 1200 also includes a graphics processor 1208 for performing graphics processing operations. In at least one embodiment, the graphics processor 1208 is coupled to a shared cache unit 1206 and a system agent core 1210 including one or more integrated memory controllers 1214. In at least one embodiment, the system agent core 1210 also includes a display controller 1211 for driving the graphics processor output to one or more coupled displays. In at least one embodiment, the display controller 1211 may also be a separate module coupled to the graphics processor 1208 via at least one interconnect, or may be integrated within the graphics processor 1208.

[0176] In at least one embodiment, a ring-based interconnect unit 1212 is used to couple the internal components of the processor 1200. In at least one embodiment, alternative interconnect units may be used, such as point-to-point interconnects, switched interconnects, or other technologies. In at least one embodiment, the graphics processor 1208 is coupled to the ring interconnect 1212 via an I / O link 1213.

[0177] In at least one embodiment, I / O link 1213 represents at least one of a variety of I / O interconnects, including packaged I / O interconnects that facilitate communication between various processor components and high-performance embedded memory modules 1218 (e.g., eDRAM modules). In at least one embodiment, each of processor cores 1202A-1202N and graphics processor 1208 uses embedded memory modules 1218 as a shared last level cache.

[0178] In at least one embodiment, the processor cores 1202A-1202N are homogeneous cores that execute a common instruction set architecture. In at least one embodiment, the processor cores 1202A-1202N are heterogeneous in terms of instruction set architecture (ISA), wherein one or more processor cores 1202A-1202N execute a common instruction set, while one or more other processor cores 1202A-1202N execute a subset or a different instruction set of the common instruction set. In at least one embodiment, the processor cores 1202A-1202N are heterogeneous in terms of microarchitecture, wherein one or more cores with relatively high power consumption are coupled with one or more power cores with lower power consumption. In at least one embodiment, the processor 1200 can be implemented on one or more chips or implemented as a SoC integrated circuit.

[0179] In at least one embodiment, some or all of the inference and / or training logic 715 may be incorporated into the processor 1200. For example, in at least one embodiment, the training and / or inference techniques described herein may use one or more ALUs embodied in Fig.12 In addition, in at least one embodiment, the inference and / or training operations described herein may use the inference and / or training operations described herein. Fig. 7A or Figure 7B In at least one embodiment, the weight parameters may be stored in on-chip or off-chip memory and / or registers (shown or not shown) that configure the ALU of the graphics processor 1200 to execute one or more of the machine learning algorithms, neural network architectures, use cases, or training techniques described herein. Such components may be used to generate facial animations from audio data, including accurate depictions of emotional states.

[0180] Virtualized computing platform

[0181] Fig.13is an example data flow diagram of a process 1300 for generating and deploying an image processing and inference pipeline according to at least one embodiment. In at least one embodiment, the process 1300 can be deployed for use with imaging devices, processing devices, and / or other device types at one or more facilities 1302. The process 1300 can be executed within a training system 1304 and / or a deployment system 1306. In at least one embodiment, the training system 1304 can be used to perform training, deployment, and implementation of machine learning models (e.g., neural networks, object detection algorithms, computer vision algorithms, etc.) for use with the deployment system 1306. In at least one embodiment, the deployment system 1306 can be configured to offload processing and computing resources in a distributed computing environment to reduce infrastructure requirements of the facility 1302. In at least one embodiment, one or more applications in the pipeline can use or call services (e.g., reasoning, visualization, computation, AI, etc.) of the deployment system 1306 during application execution.

[0182] In at least one embodiment, some applications used in the high-level processing and reasoning pipeline may use machine learning models or other AI to perform one or more processing steps. In at least one embodiment, the machine learning model may be trained at the facility 1302 using data 1308 (e.g., imaging data) generated at the facility 1302 (and stored on one or more picture archiving and communication system (PACS) servers at the facility 1302), the machine learning model may be trained using imaging or sequencing data 1308 from another one or more facilities, or a combination thereof. In at least one embodiment, the training system 1304 may be used to provide applications, services, and / or other resources to generate a working, deployable machine learning model for the deployment system 1306.

[0183] In at least one embodiment, model registry 1324 can be backed by an object store that can support version control and object metadata. In at least one embodiment, the model registry 1324 can be backed by an object store that can support version control and object metadata. Fig.14 The object store is accessed through an application programming interface (API) compatible with the cloud 1426 of the system. In at least one embodiment, the machine learning models within the model registry 1324 can be uploaded, listed, modified, or deleted by developers or partners of the system interacting with the API. In at least one embodiment, the API can provide access to methods that allow users with appropriate credentials to associate a model with an application so that the model can be executed as part of the execution of a containerized instantiation of the application.

[0184] In at least one embodiment, training pipeline 1404 ( Fig.14) may include situations where the facility 1302 is training their own machine learning model, or has an existing machine learning model that needs to be optimized or updated. In at least one embodiment, imaging data 1308 generated by an imaging device, a sequencing device, and / or other type of device may be received. In at least one embodiment, once the imaging data 1308 is received, the AI-assisted annotation 1310 may be used to help generate annotations corresponding to the imaging data 1308 to be used as ground truth data for the machine learning model. In at least one embodiment, the AI-assisted annotation 1310 may include one or more machine learning models (e.g., a convolutional neural network (CNN)) that may be trained to generate annotations corresponding to certain types of imaging data 1308 (e.g., from certain devices). In at least one embodiment, the AI-assisted annotation 1310 may then be used directly, or may be adjusted or fine-tuned using an annotation tool to generate ground truth data. In at least one embodiment, the AI-assisted annotation 1310, the labeled clinical data 1312, or a combination thereof may be used as ground truth data for training a machine learning model. In at least one embodiment, the trained machine learning model can be referred to as an output model 1316 and can be used by the deployment system 1306 as described herein.

[0185] In at least one embodiment, training pipeline 1404 ( Fig.14) may include situations where the facility 1302 requires a machine learning model for performing one or more processing tasks for one or more applications in the deployment system 1306, but the facility 1302 may not currently have such a machine learning model (or may not have a model that is optimized, efficient, or effective for this purpose). In at least one embodiment, an existing machine learning model may be selected from the model registry 1324. In at least one embodiment, the model registry 1324 may include machine learning models that are trained to perform a variety of different reasoning tasks on imaging data. In at least one embodiment, the machine learning models in the model registry 1324 may be trained on imaging data from a different facility (e.g., a facility located remotely) rather than the facility 1302. In at least one embodiment, the machine learning model may have been trained on imaging data from one location, two locations, or any number of locations. In at least one embodiment, when training on imaging data from a specific location, the training may be performed at that location, or at least in a manner that protects the confidentiality of the imaging data or restricts the transfer of the imaging data from off-site. In at least one embodiment, once a model is trained or partially trained at one location, the machine learning model may be added to the model registry 1324. In at least one embodiment, the machine learning model can then be retrained or updated at any number of other facilities, and the retrained or updated model can be used in the model registry 1324. In at least one embodiment, the machine learning model can then be selected from the model registry 1324 (and referred to as the output model 1316) and can be deployed in the deployment system 1306 to perform one or more processing tasks for one or more applications of the deployment system.

[0186] In at least one embodiment, in training pipeline 1404 ( Fig.14), may include scenarios where the facility 1302 requires a machine learning model for performing one or more processing tasks for one or more applications in the deployment system 1306, but the facility 1302 may not currently have such a machine learning model (or may not have an optimized, efficient, or effective model). In at least one embodiment, the machine learning model selected from the model registry 1324 may not be fine-tuned or optimized for the imaging data 1308 generated at the facility 1302 due to population differences, robustness of the training data used to train the machine learning model, diversity of training data anomalies, and / or other issues with the training data. In at least one embodiment, AI-assisted annotations 1310 may be used to help generate annotations corresponding to the imaging data 1308 for use as baseline truth data for training or updating the machine learning model. In at least one embodiment, labeled clinical data 1312 may be used as baseline truth data for training the machine learning model. In at least one embodiment, retraining or updating the machine learning model may be referred to as model training 1314. In at least one embodiment, model training 1314 (e.g., AI-assisted annotation 1310, labeled clinical data 1312, or a combination thereof) can be used as ground truth data for retraining or updating a machine learning model. In at least one embodiment, the trained machine learning model can be referred to as an output model 1316 and can be used by a deployment system 1306, as described herein.

[0187] In at least one embodiment, the deployment system 1306 may include software 1318, services 1320, hardware 1322, and / or other components, features, and functions. In at least one embodiment, the deployment system 1306 may include a software "stack" such that the software 1318 may be built on top of the services 1320 and may use the services 1320 to perform some or all of the processing tasks, and the services 1320 and software 1318 may be built on top of the hardware 1322 and use the hardware 1322 to perform the processing, storage, and / or other computing tasks of the deployment system. In at least one embodiment, the software 1318 may include any number of different containers, each of which may perform an instantiation of an application. In at least one embodiment, each application may perform one or more processing tasks (e.g., reasoning, object detection, feature detection, segmentation, image enhancement, calibration, etc.) in a high-level processing and reasoning pipeline. In at least one embodiment, in addition to receiving and configuring imaging data for use by each container and / or containers used by facility 1302 after processing through the pipeline, a high-level processing and reasoning pipeline can also be defined based on the selection of different containers desired or required to process imaging data 1308 (e.g., to convert the output back into a usable data type. In at least one embodiment, the combination of containers within software 1318 (e.g., which constitute a pipeline) can be referred to as a virtual instrument (as described in more detail herein), and the virtual instrument can utilize services 1320 and hardware 1322 to perform some or all of the processing tasks of an application instantiated in a container.

[0188] In at least one embodiment, the data processing pipeline can receive input data (e.g., imaging data 1308) in a specific format in response to an inference request (e.g., a request from a user of the deployment system 1306). In at least one embodiment, the input data can represent one or more images, videos, and / or other data representations generated by one or more imaging devices. In at least one embodiment, the data can be pre-processed as part of the data processing pipeline to prepare the data for processing by one or more applications. In at least one embodiment, post-processing can be performed on the output of one or more inference tasks or other processing tasks of the pipeline to prepare output data for the next application and / or prepare the output data for transmission and / or use by the user (e.g., as a response to the inference request). In at least one embodiment, the inference task can be performed by one or more machine learning models, such as a trained or deployed neural network, which can include the output model 1316 of the training system 1304.

[0189] In at least one embodiment, tasks of a data processing pipeline may be encapsulated in containers, each container representing a discrete, fully functional instantiation of an application and a virtualized computing environment capable of referencing a machine learning model. In at least one embodiment, containers or applications may be published to a private (e.g., limited access) area of ​​a container registry (described in more detail herein), and trained or deployed models may be stored in the model registry 1324 and associated with one or more applications. In at least one embodiment, an image of an application (e.g., a container image) may be available in a container registry, and once a user selects an image from a container registry for deployment in a pipeline, the image may be used to generate an instantiated container for the application for use by the user's system.

[0190] In at least one embodiment, a developer (e.g., a software developer, a clinician, a physician, etc.) can develop, publish, and store applications (e.g., as containers) for performing image processing and / or reasoning on provided data. In at least one embodiment, the development, publishing, and / or storage can be performed using a software development kit (SDK) associated with the system (e.g., to ensure that the developed applications and / or containers conform to or are compatible with the system). In at least one embodiment, the developed applications can be tested locally (e.g., at a first facility, on data from the first facility) using the SDK, which is installed as part of the system (e.g., Fig.14 The system 1400 in FIG. 14 may support at least some of the services 1320. In at least one embodiment, because a DICOM object may contain from one to hundreds of images or other data types, and because the data varies, the developer may be responsible for managing (e.g., setting up constructs for building pre-processing into the application, etc.) the extraction and preparation of the incoming data. In at least one embodiment, once validated by the system 1400 (e.g., for accuracy), the application is made available in the container registry for selection and / or implementation by the user to perform one or more processing tasks on the data at the user's facility (e.g., a second facility).

[0191] In at least one embodiment, the developer can then share the application or container over a network for use by a system (e.g., Fig.141400). In at least one embodiment, the completed and validated application or container can be stored in the container registry, and the associated machine learning model can be stored in the model registry 1324. In at least one embodiment, the requesting entity (which provides an inference or image processing request) can browse the container registry and / or the model registry 1324 to obtain applications, containers, data sets, machine learning models, etc., select the desired combination of elements to be included in the data processing pipeline, and submit the image processing request. In at least one embodiment, the request may include input data necessary to execute the request (and patient-related data in some examples), and / or may include a selection of applications and / or machine learning models to be executed when processing the request. In at least one embodiment, the request can then be passed to one or more components (e.g., a cloud) of the deployment system 1306 to perform processing of the data processing pipeline. In at least one embodiment, the processing performed by the deployment system 1306 may include referencing elements (e.g., applications, containers, models, etc.) selected from the container registry and / or the model registry 1324. In at least one embodiment, once the results are generated by the pipeline, the results may be returned to the user for reference (eg, for viewing in a viewing application suite executing locally, on a local workstation or terminal).

[0192] In at least one embodiment, to assist in processing or executing applications or containers in the pipeline, services 1320 may be utilized. Services 1320 may include computing services, artificial intelligence (AI) services, visualization services, and / or other service types. Services 1320 may provide functionality common to one or more applications in software 1318, and thus may abstract functionality into services that may be called or utilized by applications. The functionality provided by services 1320 may run dynamically and more efficiently, and may also be able to be scaled by allowing applications to process data in parallel (e.g., using Fig.14In at least one embodiment, rather than requiring that each application that shares the same functionality provided by service 1320 must have a corresponding instance of service 1320, service 1320 can be shared between and among various applications. In at least one embodiment, service 1320 may include, but is not limited to, an inference server or engine that can be used to perform detection or segmentation tasks; a model training service that provides machine learning model training and / or retraining capabilities; a data augmentation service that provides GPU-accelerated data (e.g., DICOM, RIS, CIS, REST-compliant, RPC, raw, etc.) extraction, resizing, scaling, and / or other enhancements; a visualization service that adds image rendering effects (e.g., ray tracing, rasterization, denoising, sharpening, etc.) to add realism to two-dimensional (2D) and / or three-dimensional (3D) models; and / or a virtual instrument service that provides beamforming, segmentation, reasoning, imaging, and / or support for other applications within the pipeline of a virtual instrument.

[0193] In at least one embodiment, where the service 1320 includes an AI service (e.g., an inference service), as part of the execution of an application, one or more machine learning models can be executed by calling (e.g., as an API call) an inference service (e.g., an inference server) to execute one or more machine learning models or processing thereof. In at least one embodiment, where another application includes one or more machine learning models for a segmentation task, the application can call the inference service to execute the machine learning model for performing one or more processing operations associated with the segmentation task. In at least one embodiment, the software 1318 implementing the high-level processing and inference pipeline, which includes a segmentation application and anomaly detection application, can be pipelined because each application can call the same inference service to perform one or more inference tasks.

[0194] In at least one embodiment, the hardware 1322 may include a GPU, a CPU, a graphics card, an AI / deep learning system (e.g., an AI supercomputer such as NVIDIA's DGX), a cloud platform, or a combination thereof. In at least one embodiment, different types of hardware 1322 may be used to provide efficient, purpose-built support for the software 1318 and services 1320 in the deployment system 1306. In at least one embodiment, the use of GPU processing may be implemented to perform local processing (e.g., at the facility 1302) within the AI / deep learning system, in the cloud system, and / or other processing components of the deployment system 1306 to improve the efficiency, accuracy, and effectiveness of image processing and generation. In at least one embodiment, as a non-limiting example, with respect to deep learning, machine learning, and / or high-performance computing, the software 1318 and / or services 1320 may be optimized for GPU processing. In at least one embodiment, at least some of the computing environments of the deployment system 1306 and / or training system 1304 may be executed in a data center, one or more supercomputers, or high-performance computer systems with GPU-optimized software (e.g., a combination of hardware and software of an NVIDIA DGX system). In at least one embodiment, as described herein, hardware 1322 may include any number of GPUs that may be called to perform data processing in parallel. In at least one embodiment, the cloud platform may also include GPU optimized execution for deep learning tasks, GPU processing for machine learning tasks or other computing tasks. In at least one embodiment, an AI / deep learning supercomputer and / or GPU optimized software (e.g., as provided on NVIDIA's DGX system) may be used as a hardware abstraction and scaling platform to execute a cloud platform (e.g., NVIDIA's NGC). In at least one embodiment, the cloud platform may integrate an application container cluster system or coordination system (e.g., KUBERNETES) on multiple GPUs to achieve seamless scaling and load balancing.

[0195] Fig.14 is a system diagram of an example system 1400 for generating and deploying an imaging deployment pipeline according to at least one embodiment. In at least one embodiment, the system 1400 can be used to implement Fig.13 The process 1300 and / or other processes of the system 1400 may include a high-level processing and reasoning pipeline. In at least one embodiment, the system 1400 may include a training system 1304 and a deployment system 1306. In at least one embodiment, the training system 1304 and the deployment system 1306 may be implemented using software 1318, services 1320, and / or hardware 1322, as described herein.

[0196] In at least one embodiment, system 1400 (e.g., training system 1304 and / or deployment system 1306) can be implemented in a cloud computing environment (e.g., using cloud 1426). In at least one embodiment, system 1400 can be implemented locally (with respect to a healthcare service facility) or as a combination of cloud computing resources and local computing resources. In at least one embodiment, access to APIs in cloud 1426 can be restricted to authorized users by establishing security measures or protocols. In at least one embodiment, the security protocol can include a network token, which can be signed by an authentication (e.g., AuthN, AuthZ, Gluecon, etc.) service and can carry appropriate authorization. In at least one embodiment, the API of the virtual instrument (described herein) or other instances of system 1400 can be restricted to a set of public IPs that have been audited or authorized for interaction.

[0197] In at least one embodiment, the various components of system 1400 can communicate with each other using any of a variety of different network types, including but not limited to local area networks (LANs) and / or wide area networks (WANs) via wired and / or wireless communication protocols. In at least one embodiment, communications between facilities and components of system 1400 (e.g., for sending inference requests, for receiving results of inference requests, etc.) can be transmitted via one or more data buses, wireless data protocols (Wi-Fi), wired data protocols (e.g., Ethernet), etc.

[0198] In at least one embodiment, similar to the present disclosure regarding Fig.13 As described, the training system 1304 can execute the training pipeline 1404. In at least one embodiment, where the deployment system 1306 will use one or more machine learning models in the deployment pipeline 1410, the training pipeline 1404 can be used to train or retrain one or more (e.g., pre-trained) models, and / or implement one or more pre-trained models 1406 (e.g., without retraining or updating). In at least one embodiment, as a result of the training pipeline 1404, an output model 1316 can be generated. In at least one embodiment, the training pipeline 1404 can include any number of processing steps, such as, but not limited to, conversion or adaptation of imaging data (or other input data). In at least one embodiment, different training pipelines 1404 can be used for different machine learning models used by the deployment system 1306. In at least one embodiment, similar to the description regarding Fig.13 The first example training pipeline 1404 described may be used for a first machine learning model, similar to the training pipeline 1404 described with respect to FIG. Fig.13 The second example training pipeline 1404 described may be used for a second machine learning model, similar to the one described with respect to Fig.13The training pipeline 1404 of the third example described may be used for a third machine learning model. In at least one embodiment, any combination of tasks within the training system 1304 may be used according to the requirements of each corresponding machine learning model. In at least one embodiment, one or more machine learning models may have been trained and ready for deployment, so the training system 1304 may not perform any processing on the machine learning model, and one or more machine learning models may be implemented by the deployment system 1306.

[0199] In at least one embodiment, the output model 1316 and / or the pre-trained model 1406 may include any type of machine learning model, depending on the implementation or embodiment. In at least one embodiment and without limitation, the machine learning model used by the system 1400 may include using linear regression, logistic regression, decision tree, support vector machine (SVM), naive Bayes, k-nearest neighbor (Knn), k-means clustering, random forest, dimensionality reduction algorithm, gradient boosting algorithm, neural network (e.g., autoencoder, convolution, recursion, perceptron, long / short term memory (LSTM), Hopfield, Boltzmann, deep belief, deconvolution, generative adversarial, liquid state machine, etc.), and / or other types of machine learning models.

[0200] In at least one embodiment, the labeled clinical data 1312 (e.g., traditional annotations) may be generated by any number of techniques, such as in a drawing program (e.g., an annotation program), a computer-aided design (CAD) program, a labeling program, another type of application suitable for generating ground truth annotations or labels, and / or in some examples may be hand-drawn. In at least one embodiment, the ground truth data may be synthetically generated (e.g., generated from a computer model or rendering), realistically generated (e.g., designed and generated from real-world data), machine automatically generated (e.g., using feature analysis and learning to extract features from the data and then generate labels), manually annotated (e.g., a labeler or annotation expert defines the location of the labels), and / or a combination thereof. In at least one embodiment, for each instance of imaging data 1308 (or other data type used by a machine learning model), there may be corresponding ground truth data generated by the training system 1304. In at least one embodiment, AI-assisted annotation may be performed as part of a deployment pipeline 1410; in addition to or in place of the AI-assisted annotation included in the training pipeline 1404. In at least one embodiment, the system 1400 may include a multi-layer platform that may include a software layer (e.g., software 1318) of a diagnostic application (or other application type) that may perform one or more medical imaging and diagnostic functions. In at least one embodiment, the system 1400 may be communicatively coupled (e.g., via an encrypted link) to a PACS server network of one or more facilities. In at least one embodiment, the system 1400 may be configured to access and reference data from a PACS server to perform operations such as training a machine learning model, deploying a machine learning model, image processing, reasoning, and / or other operations.

[0201] In at least one embodiment, the software layer may be implemented as a secure, encrypted, and / or authenticated API through which an application or container may be invoked (e.g., called) from an external environment (e.g., facility 1302). In at least one embodiment, the application may then call or execute one or more services 1320 to perform computational, AI, or visualization tasks associated with the respective application, and the software 1318 and / or services 1320 may utilize hardware 1322 to perform processing tasks in an effective and efficient manner.

[0202] In at least one embodiment, the deployment system 1306 can execute a deployment pipeline 1410. In at least one embodiment, the deployment pipeline 1410 can include any number of applications, which can be sequential, non-sequential, or otherwise applied to imaging data (and / or other data types) - including AI-assisted annotations, generated by imaging devices, sequencing devices, genomics devices, etc., as described above. In at least one embodiment, as described herein, the deployment pipeline 1410 for an individual device can be referred to as a virtual instrument for the device (e.g., a virtual ultrasound instrument, a virtual CT scanning instrument, a virtual sequencing instrument, etc.). In at least one embodiment, there can be more than one deployment pipeline 1410 for a single device, depending on the information desired from the data generated by the device. In at least one embodiment, there can be a first deployment pipeline 1410 in the case where an abnormality is desired to be detected from an MRI machine, and there can be a second deployment pipeline 1410 in the case where image enhancement is desired from the output of the MRI machine.

[0203] In at least one embodiment, the image generation application may include a processing task that includes the use of a machine learning model. In at least one embodiment, a user may wish to use their own machine learning model, or select a machine learning model from the model registry 1324. In at least one embodiment, a user may implement their own machine learning model or select a machine learning model to be included in an application that performs a processing task. In at least one embodiment, the application may be selectable and customizable, and by defining the construction of the application, the deployment and implementation of the application for a particular user is presented as a more seamless user experience. In at least one embodiment, by leveraging other features of the system 1400 (e.g., services 1320 and hardware 1322), the deployment pipeline 1410 may be more user-friendly, provide easier integration, and produce more accurate, efficient, and timely results.

[0204] In at least one embodiment, the deployment system 1306 may include a user interface 1414 (e.g., a graphical user interface, a web interface, etc.) that may be used to select applications to be included in the deployment pipeline 1410, to arrange applications, to modify or change applications or their parameters or configuration, to use and interact with the deployment pipeline 1410 during setup and / or deployment, and / or to otherwise interact with the deployment system 1306. In at least one embodiment, although not shown with respect to the training system 1304, the user interface 1414 (or a different user interface) may be used to select models for use in the deployment system 1306, to select models for training or retraining in the training system 1304, and / or to otherwise interact with the training system 1304.

[0205] In at least one embodiment, in addition to the application coordination system 1428, a pipeline manager 1412 may be used to manage the interaction between the application or container of the deployment pipeline 1410 and the service 1320 and / or the hardware 1322. In at least one embodiment, the pipeline manager 1412 may be configured to facilitate the interaction from application to application, from application to service 1320, and / or from application or service to hardware 1322. In at least one embodiment, although shown as included in the software 1318, this is not intended to be limiting, and in some examples, the pipeline manager 1412 may be included in the service 1320. In at least one embodiment, the application coordination system 1428 (e.g., Kubernetes, DOCKER, etc.) may include a container coordination system that can group applications into containers as a logical unit for coordination, management, scaling, and deployment. In at least one embodiment, by associating applications from the deployment pipeline 1410 (e.g., rebuilding applications, splitting applications, etc.) with individual containers, each application can be executed in a self-contained environment (e.g., at the kernel level) to improve speed and efficiency.

[0206] In at least one embodiment, each application and / or container (or its image) can be developed, modified, and deployed separately (e.g., a first user or developer can develop, modify, and deploy a first application, and a second user or developer can develop, modify, and deploy a second application separate from the first user or developer), which can allow for focus and attention on the tasks of a single application and / or container without being hindered by the tasks of another application or container. In at least one embodiment, the pipeline manager 1412 and the application coordination system 1428 can assist in communication and collaboration between different containers or applications. In at least one embodiment, as long as the expected inputs and / or outputs of each container or application are known to the system (e.g., based on the construction of the application or container), the application coordination system 1428 and / or the pipeline manager 1412 can facilitate communication and sharing of resources between and among each application or container. In at least one embodiment, since one or more applications or containers in the deployment pipeline 1410 can share the same services and resources, the application coordination system 1428 can coordinate, load balance, and determine the sharing of services or resources between and among the various applications or containers. In at least one embodiment, the scheduler can be used to track resource requirements of applications or containers, current or planned use of those resources, and resource availability. Thus, in at least one embodiment, the scheduler can allocate resources to different applications and allocate resources between and among applications, taking into account the needs and availability of the system. In some examples, the scheduler (and / or other components of the application coordination system 1428) can determine resource availability and distribution based on constraints imposed on the system (e.g., user constraints), such as quality of service (QoS), the urgency of data output (e.g., to determine whether to perform real-time processing or delayed processing), etc.

[0207] In at least one embodiment, the services 1320 utilized and shared by the applications or containers in the deployment system 1306 may include computing services 1416, AI services 1418, visualization services 1420, and / or other service types. In at least one embodiment, an application may call (e.g., execute) one or more services 1320 to perform processing operations for the application. In at least one embodiment, an application may utilize computing services 1416 to perform supercomputing or other high performance computing (HPC) tasks. In at least one embodiment, one or more computing services 1416 may be utilized to perform parallel processing (e.g., using a parallel computing platform 1430) to process data substantially simultaneously by one or more applications and / or one or more tasks of a single application. In at least one embodiment, a parallel computing platform 1430 (e.g., NVIDIA's CUDA) may implement general computing on a GPU (GPGPU) (e.g., GPU 1422). In at least one embodiment, the software layer of the parallel computing platform 1430 may provide access to a virtual instruction set and parallel computing elements of a GPU to execute computing kernels. In at least one embodiment, the parallel computing platform 1430 may include memory, and in some embodiments, memory may be shared between and among multiple containers, and / or between and among different processing tasks within a single container. In at least one embodiment, inter-process communication (IPC) calls may be generated for multiple containers and / or multiple processes within a container to enable the same data (e.g., where multiple different stages of an application or multiple applications are processing the same information) to be used for a shared memory segment from the parallel computing platform 1430. In at least one embodiment, instead of copying data and moving the data to different locations in memory (e.g., read / write operations), the same data in the same location of the memory may be used for any number of processing tasks (e.g., at the same time, at different times, etc.). In at least one embodiment, since the data as a result of processing is used to generate new data, this information of the new location of the data may be stored and shared between various applications. In at least one embodiment, the location of the data and the location of the updated or modified data may be part of the definition of how to understand the payload in the container.

[0208] In at least one embodiment, the AI ​​service 1418 may be utilized to perform a reasoning service for executing a machine learning model associated with an application (e.g., a task is to execute one or more processing tasks of an application). In at least one embodiment, the AI ​​service 1418 may utilize the AI ​​system 1424 to execute a machine learning model (e.g., a neural network such as a CNN) for segmentation, reconstruction, object detection, feature detection, classification, and / or other reasoning tasks. In at least one embodiment, the application of the deployment pipeline 1410 may use one or more output models 1316 of the self-training system 1304 and / or other models of the application to perform reasoning on the imaging data. In at least one embodiment, two or more examples of reasoning using the application coordination system 1428 (e.g., a scheduler) may be available. In at least one embodiment, the first category may include a high priority / low latency path that can achieve a higher service level agreement, such as for performing reasoning on an urgent request in an emergency, or for a radiologist during a diagnostic process. In at least one embodiment, the second category may include a standard priority path that may be used for requests that may not be urgent or where analysis can be performed at a later time. In at least one embodiment, application coordination system 1428 can allocate resources (e.g., services 1320 and / or hardware 1322) for different reasoning tasks of AI service 1418 based on priority paths.

[0209] In at least one embodiment, the shared memory may be installed to the AI ​​service 1418 in the system 1400. In at least one embodiment, the shared memory may operate as a cache (or other storage device type) and may be used to process inference requests from applications. In at least one embodiment, when an inference request is submitted, a set of API instances of the deployment system 1306 may receive the request and may select one or more instances (e.g., for best fit, for load balancing, etc.) to process the request. In at least one embodiment, in order to process the request, the request may be entered into a database, and if it is not already in the cache, the machine learning model may be located from the model registry 1324, a verification step may ensure that the appropriate machine learning model is loaded into the cache (e.g., shared storage), and / or a copy of the model may be saved to the cache. In at least one embodiment, if the application is not yet running or there are not enough instances of the application, a scheduler (e.g., a scheduler of the pipeline manager 1412) may be used to start the application referenced in the request. In at least one embodiment, if the inference server has not yet been started to execute the model, the inference server may be started. Any number of inference servers may be started for each model. In at least one embodiment, in a pull model that clusters inference servers, the model can be cached whenever load balancing is beneficial. In at least one embodiment, the inference servers can be statically loaded into the corresponding distributed servers.

[0210] In at least one embodiment, inference can be performed using an inference server running in a container. In at least one embodiment, an instance of an inference server can be associated with a model (and optionally with multiple versions of a model). In at least one embodiment, if an instance of an inference server does not exist when a request is received to perform inference on a model, a new instance can be loaded. In at least one embodiment, when the inference server is started, the model can be passed to the inference server so that the same container can be used to serve different models as long as the inference server is running as a different instance.

[0211] In at least one embodiment, during application execution, a request for reasoning for a given application may be received, and a container (e.g., an instance of a hosted reasoning server) may be loaded (if not already loaded), and a launcher may be called. In at least one embodiment, preprocessing logic in the container may load, decode, and / or perform any additional preprocessing on the incoming data (e.g., using a CPU and / or GPU). In at least one embodiment, once the data is ready for reasoning, the container may perform reasoning on the data as needed. In at least one embodiment, this may include a single reasoning call for one image (e.g., a hand X-ray), or may require reasoning for hundreds of images (e.g., a chest CT). In at least one embodiment, the application may summarize the results before completion, which may include, but is not limited to, a single confidence score, pixel-level segmentation, voxel-level segmentation, generating visualizations, or generating text to summarize the results. In at least one embodiment, different priorities may be assigned to different models or applications. For example, some models may have a real-time (TAT less than 1 minute) priority, while other models may have a lower priority (e.g., TAT less than 10 minutes). In at least one embodiment, model execution time may be measured from a requesting mechanism or entity and may include collaborative network traversal time as well as execution time of an inference service.

[0212] In at least one embodiment, the transmission of requests between the service 1320 and the reasoning application can be hidden behind a software development kit (SDK), and robust transmission can be provided through queues. In at least one embodiment, requests will be placed in queues through an API for individual application / tenant ID combinations, and the SDK will pull requests from the queue and provide the request to the application. In at least one embodiment, the name of the queue can be provided in the environment from which the SDK will pick up the queue. In at least one embodiment, asynchronous communication through queues may be useful because it can allow any instance of the application to pick up work when it is available. Results can be transmitted back through the queue to ensure that no data is lost. In at least one embodiment, the queue can also provide the ability to split the work, because the highest priority work can enter the queue connected to most instances of the application, and the lowest priority work can enter the queue connected to a single instance, which processes the tasks in the order received. In at least one embodiment, the application can run on a GPU-accelerated instance generated in the cloud 1426, and the reasoning service can perform reasoning on the GPU.

[0213] In at least one embodiment, visualization services 1420 may be utilized to generate visualizations for viewing application and / or deployment pipeline 1410 outputs. In at least one embodiment, visualization services 1420 may utilize GPU 1422 to generate visualizations. In at least one embodiment, visualization services 1420 may implement rendering effects such as ray tracing to generate higher quality visualizations. In at least one embodiment, visualizations may include, but are not limited to, 2D image rendering, 3D volume rendering, 3D volume reconstruction, 2D tomographic slices, virtual reality displays, augmented reality displays, and the like. In at least one embodiment, a virtualized environment may be used to generate a virtual interactive display or environment (e.g., a virtual environment) for system users (e.g., doctors, nurses, radiologists, etc.) to interact with. In at least one embodiment, visualization services 1420 may include internal visualizers, movies, and / or other rendering or image processing capabilities or functions (e.g., ray tracing, rasterization, internal optics, etc.).

[0214] In at least one embodiment, hardware 1322 may include GPUs 1422, AI systems 1424, cloud 1426, and / or any other hardware for executing training system 1304 and / or deployment system 1306. GPUs 1422 (e.g., NVIDIA's TESLA and / or QUADRO GPUs) may include any number of GPUs that may be used to perform processing tasks for any feature or functionality of compute services 1416, AI services 1418, visualization services 1420, other services, and / or software 1318. For example, for AI services 1418, GPUs 1422 may be used to perform pre-processing on imaging data (or other data types used by machine learning models), post-processing on the output of machine learning models, and / or perform inference (e.g., to execute machine learning models). Cloud 1426, AI systems 1424, and / or other components of system 1400 may use GPUs 1422. Cloud 1426 may include a GPU-optimized platform for deep learning tasks. AI system 1424 may use a GPU, and cloud 1426 (or at least part of a task being deep learning or reasoning) may be executed using one or more AI systems 1424. Likewise, while hardware 1322 is shown as discrete components, this is not intended to be limiting, and any component of hardware 1322 may be combined with or utilized by any other component of hardware 1322.

[0215] AI system 1424 may include a purpose-built computing system (e.g., a supercomputer or HPC) configured for reasoning, deep learning, machine learning, and / or other artificial intelligence tasks. In addition to CPU, RAM, storage, and / or other components, features, or functions, AI system 1424 (e.g., NVIDIA's DGX) may also include software (e.g., a software stack) that can use multiple GPUs 1422 to perform GPU-optimized tasks. One or more AI systems 1424 may be implemented in a cloud 1426 (e.g., in a data center) to perform some or all of the AI-based processing tasks of system 1400.

[0216] In at least one embodiment, cloud 1426 may include a GPU-accelerated infrastructure (e.g., NVIDIA's NGC) that may provide a GPU-optimized platform for executing processing tasks of system 1400. Cloud 1426 may include an AI system 1424 for executing one or more AI-based tasks of system 1400 (e.g., as a hardware abstraction and scaling platform). Cloud 1426 may be integrated with an application orchestration system 1428 that utilizes multiple GPUs to enable seamless scaling and load balancing between and among applications and services 1320. As described herein, cloud 1426 may be responsible for executing at least some services 1320 of system 1400, including compute services 1416, AI services 1418, and / or visualization services 1420. The cloud 1426 can perform large and small batch inference (e.g., executing NVIDIA's TENSORRT), provide accelerated parallel computing APIs and platforms 1430 (e.g., NVIDIA's CUDA), execute application coordination system 1428 (e.g., Kubernetes), provide graphics rendering APIs and platforms (e.g., for ray tracing, 2D graphics, 3D graphics, and / or other rendering techniques to produce higher quality movie effects), and / or can provide other functionality for system 1400.

[0217] In summary, the disclosed technology performs audio-driven facial animation with adaptive speech rate. An audio clip is analyzed to determine whether the speech rate within the audio clip exceeds one or more thresholds. For example, a neural network and / or another type of audio processing component can be used to detect the number of phonemes, the number of transitions between phonemes, and / or another measure of speech rate within a 0.5 second audio clip. The detected speech rate can be compared to one or more thresholds representing individual speech rates that are higher than a normal speech rate. If the detected speech rate reaches or exceeds a given threshold, it can be determined that the detected speech rate is equal to or higher than the high speech rate represented by the threshold. If the detected speech rate is below all thresholds, it can be determined that the detected speech rate corresponds to a normal speech rate.

[0218] When it is determined that the speech rate associated with the audio clip exceeds a given threshold, the audio features extracted from the audio clip are upsampled, and some or all of the upsampled audio features are provided as input to one or more machine learning models for performing facial animation for the audio clip. For example, an encoder and / or another type of neural network can be used to convert the audio clip into a list of 25 audio features. In the list, the features can be sorted in time and represent a continuous subset of the audio clip. When the speech rate of the audio clip exceeds a threshold representing a high speech rate, bilinear upsampling can be performed on the audio features to increase the number of audio features by a corresponding factor. A subset of the upsampled audio features can be selected and input into one or more neural networks. Then, one or more neural networks can convert the input subset of the upsampled audio features into vertex positions, geometric deformations, and / or other representations of a facial shape or expression that matches the audio clip.

[0219] One technical advantage of the disclosed technology over existing methods is the ability to adapt audio-driven facial animation techniques to audio with varying speech rates. Therefore, when compared to conventional audio-driven facial animation methods that do not account for varying speech rates in the input audio, the disclosed technology can be used to improve the accuracy, realism, and / or expressiveness of facial animation. Another advantage of the disclosed technology is that improvements in animation quality for fast speech rates can be achieved without the need to use audio data that includes varying speech rates to train machine learning models involved in audio-driven facial animation. As a result, the disclosed technology avoids the additional overhead and resource consumption associated with collecting additional training data for the machine learning model and using the additional training data to train the machine learning model.

[0220] 1. In some embodiments, a method includes: determining that a speech rate associated with an audio clip exceeds a threshold; upsampling a first feature set associated with the audio clip to a second feature set of a size different from that of the first feature set based at least on the speech rate exceeding the threshold; and generating a facial animation output corresponding to the audio clip using one or more machine learning models and based at least on at least a subset of the second feature set.

[0221] 2. The method as described in clause 1 further includes: determining that a second speech rate associated with the second audio segment exceeds a second threshold, which is higher than a threshold; upsampling a third feature set associated with the second audio segment to a fourth feature set having a size different from the second feature set based at least on the second speech rate exceeding the second threshold; and generating a second facial animation output corresponding to the second audio segment using the one or more machine learning models and based at least on at least a subset of the fourth feature set.

[0222] 3. A method as described in any of clauses 1-2, wherein the first feature set is upsampled to the second feature set based on at least a first upsampling factor associated with the threshold, and the third feature set is upsampled to the fourth feature set based on at least a second upsampling factor associated with the second threshold.

[0223] 4. The method of any one of clauses 1-3, further comprising: using an encoder neural network to convert the audio clip into a time-varying feature sequence corresponding to the first feature set.

[0224] 5. The method of any of clauses 1-4, further comprising extracting the audio segment from the audio sample at a predetermined offset from a start point of a previous audio segment extracted from the audio sample.

[0225] 6. The method of any of clauses 1-5, further comprising: executing a speech rate detection neural network to determine the speech rate associated with the audio segment.

[0226] 7. The method of any one of clauses 1-6, wherein upsampling the first feature set to the second feature set comprises: performing bilinear upsampling on the first feature set to generate the second feature set.

[0227] 8. A method as described in any of clauses 1-7, wherein generating a facial animation output comprises: extracting a subset of the second feature set from a central portion of a timeline spanned by the second feature set; executing a pronunciation neural network included in the one or more machine learning models to generate a third feature set based at least on additional inputs, the additional inputs comprising the subset of the second feature set and a fourth feature set representing an emotional state associated with the audio segment; and executing an output neural network included in the one or more machine learning models to convert the third feature set into the facial animation output.

[0228] 9. A method as described in any of clauses 1-8, wherein the speech rate corresponds to one of the number of phonemes per time period or the number of transitions between phonemes per time period.

[0229] 10. A method as described in any of clauses 1-9, wherein the facial animation output includes at least one of head displacement, jaw displacement, eye rotation, a set of skin vertex positions, a set of tongue vertex positions, or a set of principal component analysis (PCA) coefficients.

[0230] 11. In some embodiments, the processor includes one or more circuits for performing operations including: determining that a speech rate associated with an audio segment exceeds a threshold; upsampling a first feature set associated with the audio segment to a second feature set having a size different from that of the first feature set based at least on the speech rate exceeding the threshold; and generating a facial animation output corresponding to the audio segment using one or more machine learning models and based at least on at least a subset of the second feature set.

[0231] 12. A processor as described in clause 11, wherein the operation also includes: determining that a second speech rate associated with a second audio segment does not exceed a threshold; and generating a second facial animation output for the second audio segment using the one or more machine learning models and based on at least a third feature set, based at least on the second speech rate not exceeding the threshold.

[0232] 13. A processor as described in any of clauses 11-12, wherein the operation further includes: determining that a second speech rate associated with a second audio segment exceeds one or more thresholds, and the one or more thresholds are higher than a threshold; at least based on the second speech rate exceeding the one or more thresholds, upsampling a third feature set associated with the second audio segment to a fourth feature set, wherein the fourth feature set is larger than the second feature set; and generating a second facial animation output for the second audio segment using the one or more machine learning models and based at least on additional input, the additional input including at least a subset of the fourth feature set.

[0233] 14. A processor as described in any of clauses 11-13, wherein the first feature set is upsampled to the second feature set based on at least a first upsampling factor associated with the threshold, and the third feature set is upsampled to the fourth feature set based on at least a second upsampling factor associated with a highest threshold included in the one or more thresholds.

[0234] 15. A processor as described in any of clauses 11-14, wherein generating a facial animation output includes: extracting a subset of the second feature set from a central portion of a timeline spanned by the second feature set; executing a pronunciation neural network included in the one or more machine learning models to generate a third feature set based at least on additional inputs, the additional inputs comprising the subset of the second feature set and a fourth feature set representing at least one of an emotional state or a style; and executing an output neural network included in the one or more machine learning models to convert the third feature set into the facial animation output.

[0235] 16. A processor as described in any of clauses 11-15, wherein the subset of the second feature set is equal in number to the first feature set.

[0236] 17. A processor as described in any of clauses 11-16, wherein the first feature set is upsampled to the second feature set using a bilinear upsampling technique.

[0237] 18. A processor as described in any of clauses 11-17, wherein the processor is included in at least one of: a system for performing simulation operations, a system for performing digital twin operations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system implemented using an edge device; a system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content; a system implemented using a robot; a system for performing conversational AI operations; a system implemented using one or more large language models (LLMs); a system for generating synthetic data; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0238] 19. In some embodiments, a system includes one or more processing units, which are used to perform operations, including: determining that a speech rate associated with an audio segment exceeds a threshold; at least based on the speech rate exceeding the threshold, upsampling a first feature set associated with the audio segment to a second feature set of a size different from the first feature set; and using one or more machine learning models and based on at least a subset of the second feature set to generate facial animation output corresponding to the audio segment.

[0239] 20. A system as described in clause 19, wherein the system is included in at least one of the following: a system for performing simulation operations; a system for performing digital twin operations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system implemented using an edge device; a system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content; a system implemented using a robot; a system for performing conversational AI operations; a system implemented using one or more large language models (LLMs); a system for generating synthetic data; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0240] Other variations are within the spirit of the present disclosure. Thus, while the disclosed technology is susceptible to various modifications and alternative constructions, certain illustrated embodiments thereof are shown in the drawings and have been described in detail above. However, it should be understood that it is not intended to limit the present disclosure to one or more specific forms disclosed, but on the contrary, it is intended to cover all modifications, alternative constructions, and equivalents that fall within the spirit and scope of the present disclosure as defined by the appended claims.

[0241] The present disclosure may be described in the general context of computer code or machine-usable instructions (including computer executable instructions, such as program modules) being executed by a computer or other machine (such as a personal data assistant or other handheld device). In general, program modules, including routines, programs, objects, components, data structures, etc., refer to code that performs a particular task or implements a particular abstract data type. The present disclosure may be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. The present disclosure may also be implemented in a distributed computing environment in which tasks are performed by remote processing devices linked through a communications network.

[0242] As used herein, the description of "and / or" about two or more elements should be interpreted as meaning only one element, or a combination of elements. For example, "element A, element B, and / or element C" may include only element A, only element B, only element C, element A and element B, element A and element C, element B and element C, or elements A, B, and C. In addition, "at least one of element A or element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B. In addition, "at least one of element A and element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B.

[0243] The subject matter of the present disclosure is described in detail herein to meet statutory requirements. However, the description itself is not intended to limit the scope of the present disclosure. On the contrary, the inventors have anticipated that the claimed subject matter may also be embodied in other ways in conjunction with other current or future technologies to include different steps or combinations of steps similar to the steps described in this document. In addition, although the terms "step" and / or "box" may be used herein to refer to different elements of the method employed, unless and in addition to the order of the various steps being explicitly described, the terms should not be interpreted as implying any particular order among or between the various steps disclosed herein.

Claims

1. A method comprising: determining that a speech rate associated with the audio segment exceeds a threshold; upsampling a first feature set associated with the audio segment to a second feature set having a size different from the first feature set based at least on the speech rate exceeding the threshold; as well as A facial animation output corresponding to the audio clip is generated using one or more machine learning models and based at least on at least a subset of the second set of features.

2. The method of claim 1, further comprising: determining that a second speech rate associated with the second audio segment exceeds a second threshold, the second threshold being higher than the threshold; Based at least on the second speech rate exceeding the second threshold, upsampling a third feature set associated with the second audio segment to a fourth feature set having a size different from the second feature set; as well as A second facial animation output corresponding to the second audio clip is generated using the one or more machine learning models and based on at least a subset of the fourth feature set.

3. The method of claim 2, wherein the first feature set is upsampled to the second feature set based on at least a first upsampling factor associated with the threshold, and the third feature set is upsampled to the fourth feature set based on at least a second upsampling factor associated with the second threshold.

4. The method of claim 1, further comprising: The audio segment is converted into a time-varying feature sequence corresponding to the first feature set using an encoder neural network.

5. The method of claim 1, further comprising: The audio segment is extracted from the audio sample at a predetermined offset from the beginning of a previous audio segment extracted from the audio sample.

6. The method of claim 1, further comprising: A speech rate detection neural network is executed to determine the speech rate associated with the audio segment.

7. The method of claim 1, wherein upsampling the first feature set to the second feature set comprises: Bilinear upsampling is performed on the first feature set to generate the second feature set.

8. The method of claim 1, wherein generating the facial animation output comprises: extracting the subset of the second feature set from a central portion of a time axis spanned by the second feature set; executing a pronunciation neural network included in the one or more machine learning models to generate a third feature set based on at least additional input, the additional input comprising the subset of the second feature set and a fourth feature set representing an emotional state associated with the audio segment; as well as An output neural network included in the one or more machine learning models is executed to convert the third feature set into the facial animation output.

9. The method of claim 1, wherein the speech rate corresponds to one of a number of phonemes per time segment or a number of transitions between phonemes per time segment.

10. The method of claim 1, wherein the facial animation output comprises at least one of head displacement, jaw displacement, eye rotation, a set of skin vertex positions, a set of tongue vertex positions, or a set of principal component analysis (PCA) coefficients.

11. A processor, comprising: One or more circuits configured to perform operations comprising: determining that a speech rate associated with the audio segment exceeds a threshold; Based at least on the speech rate exceeding the threshold, upsampling a first feature set associated with the audio segment to a second feature set of a size different from the first feature set; and A facial animation output corresponding to the audio clip is generated using one or more machine learning models and based at least on at least a subset of the second set of features.

12. The processor of claim 11, wherein the operations further comprise: determining that a second speech rate associated with the second audio segment does not exceed the threshold; as well as Based at least on the second speech rate not exceeding the threshold, generating a second facial animation output for the second audio segment using the one or more machine learning models and based on at least a third feature set.

13. The processor of claim 11, wherein the operations further comprise: determining that a second speech rate associated with the second audio segment exceeds one or more thresholds, the one or more thresholds being higher than the threshold; Based at least on the second speech rate exceeding the one or more thresholds, upsampling a third feature set associated with the second audio segment to a fourth feature set, wherein the fourth feature set is larger than the second feature set; as well as A second facial animation output is generated for the second audio clip using the one or more machine learning models and based at least on additional inputs, the additional inputs including at least a subset of the fourth feature set.

14. A processor as claimed in claim 13, wherein the first feature set is upsampled to the second feature set based on at least a first upsampling factor associated with the threshold, and the third feature set is upsampled to the fourth feature set based on at least a second upsampling factor associated with a highest threshold included in the one or more thresholds.

15. The processor of claim 11 , wherein generating the facial animation output comprises: extracting the subset of the second feature set from a central portion of a time axis spanned by the second feature set; executing a pronunciation neural network included in the one or more machine learning models to generate a third feature set based at least on additional inputs, the additional inputs comprising the subset of the second feature set and a fourth feature set representing at least one of an emotional state or a style; as well as An output neural network included in the one or more machine learning models is executed to convert the third feature set into the facial animation output.

16. The processor of claim 15, wherein the subset of the second feature set is equal in number to the first feature set.

17. The processor of claim 11, wherein the first feature set is upsampled to the second feature set using a bilinear upsampling technique.

18. The processor of claim 11, wherein the processor is included in at least one of: A system for performing simulation operations; Systems for performing digital twin operations; A system for performing collaborative content creation of 3D assets; Systems for performing deep learning operations; Systems implemented using edge devices; A system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content; Systems implemented using robots; Systems for performing conversational AI operations; A system implemented using one or more Large Language Models (LLMs); Systems for generating synthetic data; A system comprising one or more virtual machines VM; A system implemented at least in part in a data center; or A system implemented at least in part using cloud computing resources.

19. A system comprising: One or more processing units, the one or more processing units configured to perform operations, the operations comprising: determining that a speech rate associated with the audio segment exceeds a threshold; Based at least on the speech rate exceeding the threshold, upsampling a first feature set associated with the audio segment to a second feature set of a size different from the first feature set; and A facial animation output corresponding to the audio clip is generated using one or more machine learning models and based at least on at least a subset of the second set of features.

20. The system of claim 19, wherein the system is included in at least one of: A system for performing simulation operations; Systems for performing digital twin operations; A system for performing collaborative content creation of 3D assets; Systems for performing deep learning operations; Systems implemented using edge devices; A system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content; Systems implemented using robots; Systems for performing conversational AI operations; A system implemented using one or more Large Language Models (LLMs); Systems for generating synthetic data; A system comprising one or more virtual machines VM; A system implemented at least in part in a data center; or A system implemented at least in part using cloud computing resources.

Images