Flow-Independent Neural Video Compression
By employing a fully convolutional neural network with autoencoders for motion compensation, the inefficiencies of flow-based warping in traditional neural video compression are addressed, enabling efficient and hardware-friendly video compression on mobile devices.
Patent Information
- Application Number
- JP2024529357
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-10-27
- Filing Date
- 2022-10-28
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2042-10-28
AI Technical Summary
Traditional neural video compression methods utilizing flow-based warping are inefficient for mobile devices with limited resources, as they require significant computational power and are not hardware-friendly.
Implementing a fully convolutional neural network to replace flow-based warping, using autoencoders for motion compensation in video compression, which extracts feature representations of video frames without relying on optical flow.
This approach enables efficient video compression on mobile devices by reducing computational requirements and improving hardware compatibility, while maintaining effective motion compensation.
Smart Images

Figure 2024542462000001_ABST
Abstract
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001]
[0001] This application claims priority to U.S. provisional patent application No. 63 / 284,569, entitled "FLOW-AGNOSTIC NEURAL VIDEO COMPRESSION," filed on November 30, 2021, which claims the benefit of U.S. provisional patent application No. 17 / 975,471, entitled "FLOW-AGNOSTIC NEURAL VIDEO COMPRESSION," filed on October 27, 2022, the disclosures of which are expressly incorporated by reference in their entireties into this specification. [Technical field]
[0002] Aspects of the present disclosure relate generally to neural networks, and more particularly, to video compression using artificial neural networks. [Background technology]
[0003]
[0003] An artificial neural network may comprise an interconnected group of artificial neurons (e.g., neuron models). An artificial neural network may be a computational device or may be represented as a method to be executed by a computational device. A convolutional neural network is a type of feed-forward artificial neural network. A convolutional neural network may include a collection of neurons, each with a receptive field, that collectively tile the input space. Convolutional neural networks (CNNs), such as deep convolutional neural networks (DCNs), have numerous applications. In particular, these neural network architectures are used in a variety of technologies, such as image recognition, pattern recognition, speech recognition, autonomous driving, and other classification tasks.
[0004]
[0004] Neural networks are successfully applied in multimedia compression, such as video compression and image compression. Image compression is useful for communicating multimedia, such as images and videos. However, users may have different computer architectures with different computing capabilities and may operate such devices under different network conditions. Conventional neural video compression methods utilize flow-based warping for motion compensation. However, flow-based warping may not be very efficient for mobile devices that may have limited resources. Summary of the Invention
[0005]
[0005] The present disclosure is set out in respective independent claims. Some aspects of the disclosure are set out in dependent claims.
[0006]
[0006] In one aspect of the disclosure, a processor-implemented method for video compression using an artificial neural network (ANN) is provided. The method includes receiving a video by the ANN. The method also includes extracting, by the ANN, a first set of features of a current frame of the video and a second set of features of a reference frame of the video. In addition, the method includes determining, by the ANN, an estimate of a correlation feature between the first set of features of the current frame and the second set of features of the reference frame. Further, the method includes transmitting the estimate of the correlation feature to a receiver.
[0007]
[0007] In one aspect of the disclosure, an apparatus for video compression using an artificial neural network (ANN) is provided. The apparatus includes a memory and one or more processors coupled to the memory. The processor(s) is configured to receive a video by the ANN. The processor(s) is also configured to extract, by the ANN, a first set of features of a current frame of the video and a second set of features of a reference frame of the video. In addition, the processor(s) is configured to determine, by the ANN, an estimate of a correlation feature between the first set of features of the current frame and the second set of features of the reference frame. Furthermore, the processor(s) is configured to transmit the estimate of the correlation feature to a receiver.
[0008]
[0008] In one aspect of the disclosure, a processor-implemented method for video compression using an artificial neural network (ANN) is provided. The method includes receiving, by the ANN, a reference frame of the video and an estimate of a correlation feature between a current frame of the video and the reference frame of the video. The method also includes extracting, by the ANN, a first set of features of the reference frame of the video. In addition, the method includes decoding, by the ANN, the estimate of the correlation feature. Furthermore, the method includes generating, by the ANN, a prediction of the current frame of the video based on the decoded estimate of the correlation feature and the first set of features of the reference frame of the video.
[0009]
[0009] In one aspect of the disclosure, an apparatus for video compression using an artificial neural network (ANN) is provided. The apparatus includes a memory and one or more processors coupled to the memory. The processor(s) is configured to receive, by the ANN, a reference frame of the video and an estimate of a correlation feature between a current frame of the video and the reference frame of the video. The processor(s) is also configured to extract, by the ANN, a first set of features of the reference frame of the video. In addition, the processor(s) is configured to decode, by the ANN, the estimate of the correlation feature. Furthermore, the processor(s) is configured to generate, by the ANN, a prediction of a current frame of the video based on the decoded estimate of the correlation feature and the first set of features of the reference frame of the video.
[0010]
[0010] Additional features and advantages of the present disclosure are described below. Those skilled in the art will appreciate that the present disclosure may be easily utilized as a basis for modifying or designing other structures for carrying out the same purpose of the present disclosure. Those skilled in the art will also appreciate that such equivalent constructions do not depart from the teachings of the present disclosure as set forth in the appended claims. The novel features believed to be characteristic of the present disclosure, both in terms of its organization and method of operation, together with further objects and advantages, will be better understood from the following description when considered in conjunction with the accompanying figures. It should be expressly understood, however, that each of the figures is provided for illustration and description only, and is not intended to define the scope of the present disclosure. [Brief description of the drawings]
[0011]
[0011] The features, nature, and advantages of the present disclosure will become more apparent from the detailed description set forth below when read in conjunction with the drawings in which like reference characters identify correspondingly throughout. [Figure 1]
[0012] FIG. 1 illustrates an example implementation of a neural network using a system on a chip (SOC) that includes a general-purpose processor, in accordance with some aspects of the present disclosure. [Figure 2A]
[0013] FIG. 1 illustrates a neural network according to an aspect of the present disclosure. [Figure 2B] FIG. 1 illustrates a neural network according to an aspect of the present disclosure. [Figure 2C] FIG. 1 illustrates a neural network according to an aspect of the present disclosure. [Figure 2D]
[0014] FIG. 1 illustrates an exemplary deep convolutional network (DCN), in accordance with aspects of the present disclosure. [Diagram 3]
[0015] FIG. 1 is a block diagram illustrating an example deep convolutional network (DCN), in accordance with aspects of the present disclosure. [Figure 4]
[0016] FIG. 1 is a block diagram illustrating an example software architecture that may modularize artificial intelligence (AI) functionality, according to aspects of the present disclosure. [Diagram 5]
[0017] 1 is a block diagram illustrating an example architecture for flow-independent video compression, according to aspects of the present disclosure. [Figure 6] 1 is a block diagram illustrating an example architecture for flow-independent video compression, according to aspects of the present disclosure. [Figure 7] 1 is a block diagram illustrating an example architecture for flow-independent video compression, according to aspects of the present disclosure. [Figure 8]
[0018] FIG. 1 is a flow diagram illustrating an example process for video compression using an artificial neural network, according to aspects of the present disclosure. [Figure 9] FIG. 1 is a flow diagram illustrating an example process for video compression using an artificial neural network, according to aspects of the present disclosure. [Figure 10] FIG. 1 is a flow diagram illustrating an example process for video compression using an artificial neural network, according to aspects of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0012]
[0019] The detailed description of the invention described below with reference to the accompanying drawings illustrates various configurations and does not represent the only configurations in which the described concepts may be practiced. The detailed description of the invention includes specific details intended to provide a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In some instances, well-known structures and components are shown in block diagram form to avoid obscuring such concepts.
[0013]
[0020] Based on the teachings, those skilled in the art should understand that the scope of the present disclosure is intended to encompass any aspect of the present disclosure, whether implemented independently of or in combination with any other aspect of the present disclosure. For example, an apparatus may be implemented or a method may be practiced using any number of the aspects described. In addition, the scope of the present disclosure is intended to encompass such an apparatus or method that is practiced using other structures, functions, or structures and functions in addition to or other than the various aspects of the present disclosure described. It should be understood that any aspect of the present disclosure disclosed may be embodied by one or more elements of a claim.
[0014]
[0021] The word "exemplary" is used to mean "serving as an example, instance, or illustration." Any aspect described as "exemplary" is not necessarily to be construed as preferred or advantageous over other aspects.
[0015]
[0022] Although specific aspects are described, many variations and permutations of these aspects fall within the scope of the present disclosure. Although some benefits and advantages of the preferred aspects are described, the scope of the present disclosure is not limited to any particular benefit, application, or purpose. Rather, the aspects of the present disclosure are intended to be broadly applicable to different technologies, system configurations, networks, and protocols, some of which are illustrated by way of example in the figures and the following description of the preferred aspects. The detailed description and drawings are merely illustrative of the present disclosure, rather than limiting, the scope of the present disclosure being defined by the appended claims and their equivalents.
[0016]
[0023] As described, neural networks are successfully applied in multimedia compression, such as video compression. Conventional neural video compression methods utilize flow-based warping for motion compensation. However, flow-based warping may not be hardware friendly or efficient for mobile devices that may have limited resources.
[0017]
[0024] To address these and other challenges, aspects of the present disclosure relate to flow-independent neural video compression. Instead of using flow-based warping, the warping operator may be replaced with a fully convolutional neural network. In some aspects, the fully convolutional neural network may be constructed using one or more autoencoders. An autoencoder may learn to encode unlabeled data and includes an encoder that maps an input to a code and a decoder that maps the code to a reconstruction of the input.
[0018]
[0025] 1 illustrates an exemplary implementation of a system-on-chip (SOC) 100 that may include a central processing unit (CPU) 102 or a multi-core CPU configured to operate an artificial neural network (e.g., a neural end-to-end network) to provide flow-independent video compression. Variables (e.g., neural signals and synaptic weights), system parameters associated with a computational device (e.g., a neural network with weights), delays, frequency bin information, and task information may be stored in a memory block associated with a neural processing unit (NPU) 108, in a memory block associated with the CPU 102, in a memory block associated with a graphics processing unit (GPU) 104, in a memory block associated with a digital signal processor (DSP) 106, in a memory block 118, or distributed across multiple blocks. Instructions executed in the CPU 102 may be loaded from a program memory associated with the CPU 102 or from the memory block 118.
[0019]
[0026] The SOC 100 may also include a GPU 104, a DSP 106, a connectivity block 110 that may include fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, etc., as well as additional processing blocks adapted to specific functions, such as a multimedia processor 112 that may detect and recognize gestures. In one implementation, the NPU 108 is implemented in the CPU 102, the DSP 106, and / or the GPU 104. The SOC 100 may also include a sensor processor 114, image signal processors (ISPs) 116, and / or a navigation module 120 that may include a global positioning system.
[0020]
[0027] The SOC 100 may be based on an ARM instruction set. In an aspect of the present disclosure, the instructions loaded into the general purpose processor 102 may include code for receiving a video by an artificial neural network (ANN). The general purpose processor 102 may also include code for extracting, via the ANN, a first set of features of a current frame of the video and a second set of features of a reference frame of the video. In addition, the general purpose processor 102 may include code for determining, via the ANN, an estimate of a correlation feature between the first set of features of the current frame and the second set of features of the reference frame. The general purpose processor 102 may further include code for transmitting the estimate of the correlation feature to a receiver.
[0021]
[0028] In another aspect of the disclosure, the instructions loaded into the general purpose processor 102 may include code for receiving, by an artificial neural network (ANN), a reference frame of the video and an estimate of correlation features between the current frame of the video and the reference frame of the video. The general purpose processor 102 may also include code for extracting, by the ANN, a first set of features of the reference frame of the video. In addition, the general purpose processor 102 may include code for decoding, by the ANN, the estimate of the correlation features. The general purpose processor 102 may further include code for generating, by the ANN, a prediction of the current frame of the video based on the decoded estimate of the correlation features and the first set of features of the reference frame of the video.
[0022]
[0029] In a further aspect of the disclosure, the instructions loaded into the general purpose processor 102 may include code for receiving, by the ANN, a history of a plurality of frames of a video. The general purpose processor 102 may also include code for generating, by the ANN, a reference frame from the history of the plurality of frames. In addition, the general purpose processor 102 may include code for generating, by the ANN, a prediction of a current frame of the video based on the reference frame and the current frame. The general purpose processor 102 may further include code for generating a reconstruction of the current frame based on a prediction of the current frame of the video and a residual calculated based on a difference between the current frame of the video and the prediction of the current frame of the video.
[0023]
[0030] Deep learning architectures may perform object recognition tasks by learning to represent inputs at successively higher levels of abstraction within each layer, thereby building useful feature representations of the input data. In this way, deep learning addresses a major bottleneck of traditional machine learning. Prior to the advent of deep learning, machine learning approaches to object recognition problems may have relied heavily on human-designed features, possibly in combination with shallow classifiers. A shallow classifier may be, for example, a two-class linear classifier that can compare a weighted sum of feature vector components to a threshold to predict which class an input belongs to. Human-designed features may be templates or kernels that are adapted to a particular problem domain by an engineer with domain expertise. In contrast, deep learning architectures may learn, but through training, to represent features that are similar to those that a human engineer could design. Furthermore, deep networks may learn to represent and recognize new types of features that humans may not have thought of.
[0024]
[0031] A deep learning architecture may learn a hierarchy of features. When presented with visual data, for example, a first layer may learn to recognize relatively simple features such as edges in the input stream. In another example, when presented with auditory data, the first layer may learn to recognize spectral power at specific frequencies. A second layer, taking as input the output of the first layer, may learn to recognize combinations of features such as simple shapes in the case of visual data, or combinations of sounds in the case of auditory data. For example, higher layers may learn to represent complex shapes in visual data or words in auditory data. Even higher layers may learn to recognize common visual objects or spoken phrases.
[0025]
[0032] Deep learning architectures may work particularly well when applied to problems that have a natural hierarchical structure. For example, classification of electric vehicles may benefit from first learning to recognize wheels, windshields, and other features. These features may be combined in different ways at higher layers to recognize cars, trucks, and planes.
[0026]
[0033] Neural networks may be designed with various connectivity patterns. In feedforward networks, each neuron in a given layer communicates with neurons in a higher layer, so that information is passed from lower layers to higher layers. As described above, hierarchical representations may be constructed in successive layers of a feedforward network. Neural networks may also have recurrent or feedback (also called top-down) connections. In recurrent connections, the output from a neuron in a given layer may be transmitted to another neuron in the same layer. Recurrent architectures may be useful in recognizing patterns across two or more of the input data chunks delivered in a sequence to the neural network. Connections from neurons in a given layer to neurons in a lower layer are called feedback (or top-down) connections. Networks with many feedback connections may be useful when recognition of high-level concepts can help distinguish certain low-level features of the input.
[0027]
[0034] The connections between layers of a neural network may be fully connected or locally connected. FIG. 2A shows an example of a fully connected neural network 202. In the fully connected neural network 202, a neuron in a first layer may transmit its output to every neuron in a second layer, so that each neuron in the second layer receives input from every neuron in the first layer. FIG. 2B shows an example of a locally connected neural network 204. In the locally connected neural network 204, a neuron in a first layer may be connected to a limited number of neurons in the second layer. More generally, the locally connected layers of the locally connected neural network 204 may be configured such that each neuron in a layer has the same or similar connectivity pattern, but with connection strengths that may have different values (e.g., 210, 212, 214, and 216). Because higher layer neurons in a given region may receive inputs that are tuned through training to the properties of a limited subset of all inputs to the network, the connectivity patterns of local connections may give rise to spatially distinct receptive fields within the higher layers.
[0028]
[0035] One example of a locally connected neural network is a convolutional neural network. Figure 2C shows an example of a convolutional neural network 206. The convolutional neural network 206 may be configured such that the connection strengths associated with the inputs for each neuron in the second layer are shared (e.g., 208). Convolutional neural networks may be suitable for problems where the spatial location of the inputs is meaningful.
[0029]
[0036] One type of convolutional neural network is the deep convolutional network (DCN). Figure 2D shows a detailed example of a DCN 200 designed to recognize visual features from images 226 input from an image capture device 230, such as an on-board camera. The DCN 200 in this example may be trained to identify traffic signs and numbers on traffic signs. Of course, the DCN 200 may be trained for other tasks, such as identifying lane markings or identifying traffic signals.
[0030]
[0037] The DCN 200 may be trained using supervised learning. During training, the DCN 200 may be presented with an image, such as the image 226 of a speed limit sign, and then a forward pass may be computed to generate the output 222. The DCN 200 may include a feature extraction section and a classification section. Upon receiving the image 226, the convolution layer 232 may apply a convolution kernel (not shown) to the image 226 to generate the first set of feature maps 218. As an example, the convolution kernel for the convolution layer 232 may be a 5×5 kernel that generates a 28×28 feature map. In this example, four different convolution kernels were applied to the image 226 in the convolution layer 232 because four different feature maps are generated in the first set of feature maps 218. The convolution kernel may also be referred to as a filter or a convolution filter.
[0031]
[0038] The first set of feature maps 218 may be subsampled by a max pooling layer (not shown) to generate a second set of feature maps 220. The max pooling layer reduces the size of the first set of feature maps 218. That is, the size of the second set of feature maps 220, such as 14×14, is smaller than the size of the first set of feature maps 218, such as 28×28. The reduced size provides similar information to subsequent layers while reducing memory consumption. The second set of feature maps 220 may be further convolved through one or more subsequent convolutional layers (not shown) to generate one or more subsequent sets of feature maps (not shown).
[0032]
[0039] In the example of Figure 2D, the second set of feature maps 220 are convolved to generate a first feature vector 224. Furthermore, the first feature vector 224 is further convolved to generate a second feature vector 228. Each feature in the second feature vector 228 may include a number corresponding to a possible feature of the image 226, such as "sign", "60", and "100". A softmax function (not shown) may convert the numbers in the second feature vector 228 into probabilities. Thus, the output 222 of the DCN 200 may be a probability that the image 226 contains one or more features.
[0033]
[0040] In this example, the probability in output 222 for "sign" and "60" is higher than the probability for other of the outputs 222, such as "30", "40", "50", "70", "80", "90", and "100". Before training, the output 222 generated by DCN 200 may be inaccurate. Thus, an error may be calculated between the output 222 and a target output. The target output is the ground truth of image 226 (e.g., "sign" and "60"). The weights of DCN 200 may then be adjusted so that the output 222 of DCN 200 is more closely aligned with the target output.
[0034]
[0041] To adjust the weights, the learning algorithm may calculate a gradient vector for the weights. The gradient can indicate the amount by which the error would increase or decrease if the weights were adjusted. In the top layer, the gradient can correspond directly to the values of the weights connecting the activated neurons in the penultimate layer to the neurons in the output layer. In lower layers, the gradient can depend on the values of the weights and the calculated error gradients of the upper layers. The weights can then be adjusted to reduce the error. This method of adjusting weights is sometimes called "backpropagation" because it involves a "backward pass" through the neural network.
[0035]
[0042] In practice, the error gradient of the weights may be calculated over a small number of examples such that the calculated gradient approximates the true error gradient. This approximation may be referred to as stochastic gradient descent. Stochastic gradient descent may be repeated until the achievable error rate of the entire system stops decreasing or until the error rate reaches a target level. After training, the DCN 200 may be presented with new images and a forward pass through the DCN 200 may result in an output 222, which may be considered the inference or prediction of the DCN 200.
[0036]
[0043] Deep belief networks (DBNs) are probabilistic models with multiple layers of hidden nodes. DBNs may be used to extract a hierarchical representation of a training dataset. DBNs may be obtained by stacking layers of Restricted Boltzmann Machines (RBMs). RBMs are a type of artificial neural network that can learn a probability distribution over a set of inputs. RBMs are frequently used in unsupervised learning because they can learn a probability distribution without information about the class into which each input should be categorized. Using a hybrid unsupervised-supervised paradigm, the lower RBM of the DBN can be trained in an unsupervised manner and can function as a feature extractor, and the upper RBM can be trained in a supervised manner (on the joint distribution of inputs from previous layers and the target class) and can function as a classifier.
[0037]
[0044] Deep convolutional networks (DCNs) are networks of convolutional networks, composed of additional pooling and normalization layers. DCNs have achieved state-of-the-art performance for many tasks. DCNs can be trained using supervised learning, where both the input and output targets are known for a large number of examples and are used to modify the network weights by using gradient descent.
[0038]
[0045] The DCN may be a feedforward network. In addition, as described above, connections from neurons in a first layer of the DCN to groups of neurons in the next higher layer are shared across neurons in the first layer. The feedforward and shared connections of the DCN can be exploited for high speed processing. The computational burden of the DCN may be much smaller than that of a similarly sized neural network that includes, for example, recurrent or feedback connections.
[0039]
[0046] The processing of each layer of the convolutional network may be considered as a spatially invariant template or basis projection. If the input is initially decomposed into multiple channels, such as the red, green, and blue channels of a color image, the convolutional network trained on that input may be considered as three-dimensional, with two spatial dimensions along the axes of the image and a third dimension capturing color information. The output of the convolutional connections may be considered to form a feature map in the subsequent layer, with each element of the feature map (e.g., 220) receiving input from a range of neurons in the previous layer (e.g., feature map 218) and from each of multiple channels. The values in the feature map may be further processed using nonlinearities such as rectification, max(0,x), etc. Values from neighboring neurons may be further pooled, corresponding to downsampling, to provide additional local invariance and dimensionality reduction. Normalization, corresponding to whitening, may also be applied through lateral inhibition between neurons in the feature map.
[0040]
[0047] The performance of deep learning architectures can improve as more labeled data points become available or as computational power increases. Modern deep neural networks are routinely trained with computational resources thousands of times greater than those available to a typical researcher only 15 years ago. New architectures and training paradigms can further improve deep learning performance. Rectified linear units may reduce the training problem known as vanishing gradients. New training techniques may reduce overfitting and therefore allow larger models to achieve better generalization. Encapsulation techniques may extract data within a given receptive field, further improving overall performance.
[0041]
[0048] FIG. 3 is a block diagram illustrating a DCN 350. The deep convolutional network 350 may include multiple different types of layers based on connectivity and weight sharing. As shown in FIG. 3, the deep convolutional network 350 includes convolutional blocks 354A, 354B. Each of the convolutional blocks 354A, 354B may be configured with a convolutional layer (CONV) 356, a normalization layer (LNorm) 358, and a max pooling layer (MAX POOL) 360. Although only two of the convolutional blocks 354A, 354B are shown, the present disclosure is not so limited and instead any number of convolutional blocks 354A, 354B may be included in the deep convolutional network 350 according to design preference.
[0042]
[0049] The convolution layer 356 may include one or more convolution filters that may be applied to the input data to generate feature maps. The normalization layer 358 may normalize the output of the convolution filters. For example, the normalization layer 358 may provide whitening or lateral inhibition. The max pooling layer 360 may provide downsampling aggregation across the space for local invariance and dimensionality reduction.
[0043]
[0050] For example, a parallel filter bank of a deep convolutional network may be loaded onto the CPU 102 or GPU 104 of SOC 100 (e.g., FIG. 1) to achieve high performance and low power consumption. In alternative embodiments, the parallel filter bank may be loaded onto the DSP 106 or ISP 116 of SOC 100. In addition, deep convolutional network 350 may access other processing blocks that may be present on SOC 100, such as sensor processor 114 and navigation module 120, dedicated to sensors and navigation, respectively.
[0044]
[0051] The deep convolutional network 350 may also include one or more fully connected layers 362 (FC1 and FC2). The deep convolutional network 350 may further include a logistic regression (LR) layer 364. Between each layer 356, 358, 360, 362, 364 of the deep convolutional network 350, there are weights (not shown) that are to be updated. The output of each of the layers (e.g., 356, 358, 360, 362, 364) may serve as an input for a subsequent one of the layers (e.g., 356, 358, 360, 362, 364) in the deep convolutional network 350 to learn a hierarchical feature representation from the input data 352 (e.g., image, audio, video, sensor data, and / or other input data) provided initially to the convolution block 354A. The output of the deep convolutional network 350 is a classification score 366 for the input data 352. The classification score 366 may be a set of probabilities, where each probability is the probability that the input data contains a feature from the set of features.
[0045]
[0052] 4 is a block diagram illustrating an example software architecture 400 that may modularize artificial intelligence (AI) functions. Using this architecture 400, applications may be designed that may cause various processing blocks (e.g., CPU 422, DSP 424, GPU 426, and / or NPU 428) of a SOC 420 (which may be similar to SoC 100 of FIG. 1) to support adaptive rounding as disclosed for post-training quantization for an AI application 402 according to aspects of the present disclosure.
[0046]
[0053] The AI application 402 may be configured to invoke functions defined in user space 404, which may, for example, perform scene detection and recognition indicative of the location in which the device is currently operating. The AI application 402 may configure the microphone and camera differently depending on, for example, whether the scene to be recognized is an office, an auditorium, a restaurant, or an outdoor setting such as a lake. The AI application 402 may make a request to compiled program code associated with a library defined in an AI function application programming interface (API), for example, SceneDetect API 406, to provide an estimate of the current scene. The request may ultimately rely on the output of, for example, a deep neural network configured to provide an inference response based on the video and positioning data. The deep neural network may, for example, be a differential neural network configured to provide a scene estimate based on the video and positioning data.
[0047]
[0054] A runtime engine 408, which may be compiled code of a runtime framework, may be further accessible to the AI application 402. The AI application 402 may cause the runtime engine 408 to request inference, such as scene estimation, for example, at specific time intervals or triggered by an event detected by the user interface of the application 402. When caused to provide an inference response (e.g., to estimate a scene), the runtime engine 408 may in turn signal an operating system in the operating system (OS) space 410, such as kernel 412, running on the SOC 420. The operating system may then cause successive relaxations of quantization to be performed on the CPU 422, DSP 424, GPU 426, NPU 428, or some combination thereof. The CPU 422 may be accessed directly by the operating system, and the other processing blocks may be accessed through drivers, such as drivers 414, 416, or 418 for the DSP 424, GPU 426, or NPU 428, respectively. In an illustrative example, the deep neural network may be configured to run on a combination of processing blocks such as CPU 422, DSP 424, and GPU 426, or may run on NPU 428.
[0048]
[0055]
[0049]
[0056] As mentioned above, aspects of the present disclosure relate to flow-independent video compression. According to aspects of the present disclosure, instead of using a flow warping operator, motion compensation is provided via a fully convolutional operator. For example, instead of warping the previous frame using optical flow, a convolutional neural network may extract a feature representation of the video frame. Correlation features between the current frame and the previous frame may be determined. The correlation features may be encoded and transmitted to a receiver. The encoded correlation features may be decoded to generate a prediction of the current frame.
[0050]
number
[0051] This can be used together with features of the previous frame to determine
[0052]
[0057] 5 is a block diagram illustrating an example architecture 500 for flow-independent video compression in accordance with an aspect of this disclosure. The architecture 500 may include a motion network 502 and a residual network 504. The motion network 502 and the residual network 504 may be implemented, for example, as an autoencoder.
[0053]
[0058] The motion network 502 may provide motion compensation as a feature transformation. In the example of FIG. 5, the motion network 502 may include an encoder 506, a decoder 508, and a prior network 510. The encoder 506 includes a feature extractor g a 514a. a 514a may be, for example, a convolutional neural network (e.g., 350 shown in FIG. 3). a 514a may receive a video as an input. The video may include multiple frames. As shown, the feature extractor g a 514a indicates the current frame x of the video to be compressed. i The feature extractor g a 514a converts the current frame x i , extract a set of features, and compute the current frame x i In addition, the feature extractor g a 514b is the previous frame of the video
[0054]
number
[0055] FIG. 5 shows a number of feature extractors g aHowever, this is merely an example and is for ease of illustration only. Instead, one or more feature extractors g a A feature extractor g may be used. a 514b performs a step of converting a previous frame of the video through one or more convolutional layers.
[0056]
number
[0057] , extract a set of features, and
[0058]
number
[0059] The feature representations of the current and previous frames can be generated by the feature extractor g a 514a, 514b. The feature representations of the current and previous frames are concatenated via a concatenation layer 516a and fed to a feature correlation estimator g c 518.
[0060] Feature correlation estimator g c 518 may also be a convolutional neural network. c 518 is the current frame x i and the previous frame
[0061]
number
[0062] Correlation characteristics between
[0063]
number
[0064] A set of correlation features can be learned and encoded.
[0065]
number
[0066] is the current frame x i and the previous frame
[0067]
number
[0068] It can accommodate movements between.
[0069]
[0060] Correlation Features
[0070]
number
[0071] may be entropy coded via an arithmetic encoder (AE) and transmitted to the receiver. The entropy coding and decoding may be performed based on the distribution of latent correlation features learned via the prior network 510.
[0072] The decoder 508 generates the correlation feature
[0073]
number
[0074] The entropy coded latent representation of the correlation feature
[0075]
number
[0076] The entropy-encoded latent representation of is decoded via an arithmetic decoder (AD) and the code-to-feature transformer g d 520. The code-to-feature converter g d 520 is a correlation feature
[0077]
number
[0078] We can recover a latent feature representation of
[0079] The decoder 508 may also receive a reference frame 522. The reference frame 522 may be, for example, a previous frame
[0080]
number
[0081] The reference frame 522 may be subjected to a feature extractor g which extracts a set of features and generates a feature representation of the reference frame 522. a Supplied to 514c.
[0082]
[0063] Code-feature converter g d Correlation features output by 520
[0083]
number
[0084] The latent feature representation of is then fed to the reference frame 522 via the concatenation layer 516b.
[0085]
number
[0086] The feature representation of the frame is concatenated with the frame decoder g s524. The frame decoder g s 524 is the correlation feature and reference frame 522
[0087]
number
[0088] , and processing the concatenated feature representations of the current frame, which are generated via the motion network 502
[0089]
number
[0090] may be generated.
[0091] The residual network 504 is a residual encoder r a 526, Residual Decoder s 528, and the residual preceding network 512. a 526 may receive as input the residual, which is the actual current frame x i and the current frame prediction
[0092]
number
[0093] The residual encoder r can be given by the difference a 526 may extract a feature representation of the residual.
[0094] The residual feature representation is entropy coded and then input to a receiver (e.g., a residual decoder r s 528). Entropy encoding and decoding may be performed based on the distribution of latent residual features learned via prior network 512.
[0095]
[0066] Residual decoder r s528 decodes the residual feature representation and (for example,
[0096]
number
[0097] Prediction
[0098]
number
[0099] This residual prediction
[0100]
number
[0101] The current frame prediction is calculated via summation block 530.
[0102]
number
[0103] is added to the current frame x i Restoration of
[0104]
number
[0105] The current frame x i Restoration of
[0106]
number
[0107] is the next frame of the video (e.g., x i+1), may be transmitted to a receiver (eg, encoder 506 of motion network 502) for use in motion compensation for the
[0108] In some aspects, the architecture 500 may use a full convolution operator, such that motion compensation may be determined without using traditional motion compensation approaches that use optical flow.
[0109]
[0068] Figure 6 is a block diagram illustrating an example architecture 600 for flow-independent video compression according to an aspect of the present disclosure. With reference to Figure 6, the input for the architecture 500 of Figure 5 can be modified to further exploit motion information from the history of multiple previously decoded frames. In Figure 6, two previously decoded frames
[0110]
number
[0111] is shown, the disclosure is not so limited. Alternatively, two or more previously decoded frames may be included in the history of previously decoded frames. The architecture 600 may further include an extrapolator network (E) 602. The extrapolator network 602 may be, for example, a fully convolutional network. The extrapolator network 602 may be configured to extract multiple previously decoded frames from the history of previously decoded frames.
[0112]
number
[0113] The extrapolator network 602 may receive as input a previously decoded frame,
[0114]
number
[0115] Processing the reference frame prediction
[0116]
number
[0117] Reference frame prediction
[0118]
number
[0119] may be treated as an intermediate frame and input to the motion network 502 of the architecture 500. By using multiple previously decoded frames as references instead of a single previously decoded frame (e.g., as shown in FIG. 5), the accuracy of predicting the current frame may be increased.
[0120]
[0069] Figure 7 is a block diagram illustrating an example architecture 700 for flow-independent video compression according to an aspect of the present disclosure. Figure 7 is a simplified version of Figure 5 designed for efficiency optimization. With reference to Figure 7, the architecture 700 can be simplified by removing the motion network 502. Hence, the architecture 700 includes an extrapolator network 602 and a residual network 504. The extrapolator network 602 derives the current frame x from a history of previous frames (e.g., two or more frames). i Prediction
[0121]
number
[0122] The residual network 504 can directly determine the prediction
[0123]
number
[0124] and the current frame x i This is the current frame prediction.
[0125]
number
[0126] is added to the current frame x i Restoration of
[0127]
number
[0128] may be generated.
[0129]
[0070] Figure 8 is a flow diagram illustrating an example process 800 for video compression using an artificial neural network (ANN), according to an aspect of the present disclosure. As shown in Figure 8, in block 802, the process 800 receives a video via an ANN. As described with reference to Figure 5, a feature extractor g a 514a may receive a video as an input. The video may include multiple frames.
[0130] At block 804, the process 800 extracts a first set of features for the current frame of the video and a second set of features for the reference frame of the video using an ANN. a 514a converts the current frame x i , extract a set of features, and compute the current frame x i Optionally, in block 806, a feature extractor g a 514b is the previous frame of the video
[0131]
number
[0132] The feature extractor g a 514b performs a step of converting a previous frame of the video through one or more convolutional layers.
[0133]
number
[0134] , extract a set of features, and
[0135]
number
[0136] A feature representation of the
[0137] At block 808, the process 800 determines, by the ANN, an estimate of correlation features between the first set of features of the current frame and the second set of features of the reference frame. As described with reference to FIG. 5, the feature representations of the current and previous frames are respectively fed to a feature extractor g a 514a, 514b. The feature representations of the current and previous frames are concatenated via a concatenation layer 516a and fed to a feature correlation estimator g c 518. The feature correlation estimator g c 518 may also be a convolutional neural network. c 518 is the current frame x i and the previous frame
[0138]
number
[0139] Correlation characteristics between
[0140]
number
[0141] A set of correlation features can be learned and encoded.
[0142]
number
[0143] is the current frame x i and the previous frame
[0144]
number
[0145] It can accommodate movements between.
[0146] At block 810, the process 800 transmits the estimate of the correlation feature to the receiver. As described with reference to FIG.
[0147]
number
[0148] may be entropy coded via an arithmetic encoder (AE) and transmitted to a receiver (eg, decoder 508).
[0149]
[0074] Figure 9 is a flow diagram illustrating an example process 900 for video compression using an artificial neural network (ANN) according to an aspect of the present disclosure. As shown in Figure 9, at block 902, the process 900 receives a reference frame of the video by the ANN. At block 904, the process 900 receives an estimate of a correlation feature between a current frame of the video and the reference frame of the video by the ANN. As shown in Figure 5, the decoder 508 receives the reference frame of the video 522 and the current frame of the video x i and the video reference frame 522
[0150]
number
[0151] Correlation characteristics between
[0152]
number
[0153] Receive an estimate of.
[0154] At block 906, the process 900 extracts, by the ANN, a first set of features of a reference frame of the video. For example, as described with reference to FIG. 5, the decoder 508 may also receive a reference frame 522. The reference frame 522 may be, for example, a previous frame
[0155]
number
[0156] The reference frame 522 may be subjected to a feature extractor g which extracts a set of features and generates a feature representation of the reference frame 522. a Supplied to 514c.
[0157] At block 908, the process 900 decodes the estimate of the correlation feature by the ANN. As shown in FIG. 5, the decoder 508 decodes the correlation feature.
[0158]
number
[0159] The entropy coded latent representation of the correlation feature
[0160]
number
[0161] The entropy-coded latent representation of is decoded via an arithmetic decoder (AD) and a code-to-feature transformer g d 520. The code-to-feature converter g d 520 is a correlation feature
[0162]
number
[0163] We can recover a latent feature representation of
[0164] At block 910, the process 900 generates, by the ANN, a prediction of the current frame of the video based on the decoded estimate of the correlation feature and a first set of features of the reference frame of the video. As described with reference to FIG. 5, the correlation feature
[0165]
number
[0166] The latent feature representation of is concatenated with the feature representation of the reference frame 522 via a concatenation layer 516b, and the frame decoder g s 524. The frame decoder g s 524 is a correlation feature
[0167]
number
[0168] and the concatenated feature representation of the reference frame 522, and the current frame x i Prediction
[0169]
number
[0170] Generate.
[0171]
[0078] Figure 10 is a flow diagram illustrating an example process 1000 for video compression using an artificial neural network (ANN) according to an aspect of the present disclosure. As shown in Figure 10, at block 1002, the process 1000 receives, by the ANN, a history of multiple frames of a video. As described with reference to Figure 6, the architecture 600 may further include an extrapolator network (E) 602. The extrapolator network 602 may be, for example, a fully convolutional network. The extrapolator network 602 may be a multiple previously decoded frames.
[0172]
number
[0173] may receive as input:
[0174] At block 1004, the process 1000 generates a reference frame by the ANN based on the history of multiple frames of the video. For example, as described with reference to FIG. 6, the extrapolator network 602 may generate a reference frame based on a history of previously decoded frames.
[0175]
number
[0176] Processing the reference frame prediction
[0177]
number
[0178] may be generated.
[0179] At block 1006, the process 1000 generates, by the ANN, a prediction of the current frame of the video based on the reference frame and the current frame of the video. As shown in FIG. 6, the reference frame prediction created by the extrapolator network 602
[0180]
number
[0181] and the current frame x i Prediction of the current frame using
[0182]
number
[0183] is generated.
[0184] At block 1008, the process 1000 generates a reconstruction of the current frame based on a prediction of the current frame of the video and a residual calculated based on the difference between the current frame of the video and the prediction of the current frame of the video by the ANN. As shown in FIG. 6, the residual network 504 generates a reconstruction of the current frame x based on the prediction and the current frame x. i This is the current frame prediction.
[0185]
number
[0186] is added to the current frame x i Restoration of
[0187]
number
[0188] may be generated.
[0189]
[0082] Implementation examples are described in the following numbered clauses: 1. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving a video by an ANN; extracting, by an ANN, a first set of features of a current frame of the video and a second set of features of a reference frame of the video; determining, by the ANN, an estimate of a correlation feature between a first set of features of the current frame and a second set of features of the reference frame; transmitting the correlation feature estimate to a receiver; 4. A processor-implemented method comprising: 2. The processor-implemented method of claim 1, further comprising determining an estimate of the correlation features via a full convolution operator of the ANN. 3. A processor-implemented method as described in clause 1 or 2, wherein the reference frame comprises a reconstruction of a previous frame of the video. 4. The processor-implemented method of any one of clauses 1-3, wherein the reference frame is predicted based on reconstruction of multiple previous frames of the video. 5. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving, by the ANN, a reference frame of the video; receiving, by the ANN, an estimate of a correlation feature between a current frame of the video and a reference frame of the video; Extracting a first set of features of a reference frame of the video by an ANN; Decoding the correlation feature estimates by the ANN; and generating, by the ANN, a prediction for a current frame of the video based on the decoded estimate of the correlation feature and a first set of features of a reference frame of the video; 4. A processor-implemented method comprising: 6. calculating a residual based on a difference between a current frame of the video and a prediction of the current frame of the video; generating a reconstruction of the current frame of the video based on the prediction and the residual of the current frame; 6. The processor-implemented method of claim 5, further comprising: 7. The processor-implemented method of any one of clauses 5 to 6, wherein the reference frame is predicted based on reconstruction of multiple previous frames of the video. 8. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving, by the ANN, a history of a number of frames of a video; generating a reference frame from a history of multiple frames of the video using an ANN; generating, by the ANN, a prediction of a current frame of the video based on the reference frame and the current frame of the video; generating, by the ANN, a reconstruction of the current frame based on a prediction of the current frame of the video and a residual calculated based on a difference between the current frame of the video and the prediction of the current frame of the video; 4. A processor-implemented method comprising: 9. The processor-implemented method of claim 8, wherein the history includes a plurality of previously decoded frames. 10. The processor-implemented method of any one of clauses 8 to 9, further comprising encoding a difference between the current frame of video and a prediction of the current frame of video, and adding the encoding to the prediction of the current frame. 11. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: ANN receives video, extracting, by the ANN, a first set of features for a current frame of the video and a second set of features for a reference frame of the video; determining, by the ANN, an estimate of correlation features between a first set of features of the current frame and a second set of features of the reference frame; The apparatus is configured to transmit the estimate of the correlation feature to a receiver. 12. At least one processor: 12. The apparatus of claim 11, further configured to determine an estimate of the correlation feature via a full convolution operator of the ANN. 13. The apparatus of clause 11 or 12, wherein the reference frame comprises a reconstruction of a previous frame of the video. 14. The apparatus of any one of clauses 11 to 13, wherein at least one processor is further configured to predict the reference frame based on reconstruction of multiple previous frames of the video. 15. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: receiving a reference frame of the video; receiving an estimate of a correlation feature between a current frame of the video and a reference frame of the video; Extracting a first set of features for a reference frame of the video; Decoding the correlation feature estimate; The apparatus is configured to generate a prediction for a current frame of the video based on a decoded estimate of the correlation feature and a first set of features of a reference frame of the video. 16. At least one processor: Calculating a residual based on a difference between a current frame of the video and a prediction of the current frame of the video; 16. The apparatus of clause 15, further configured to generate a reconstruction of the current frame of the video based on the prediction and the residual of the current frame. 17. The apparatus of clause 15 or 16, wherein the at least one processor is further configured to predict the reference frame based on reconstruction of multiple previous frames of the video. 18. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: The ANN receives a history of multiple frames of a video; The ANN generates a reference frame from the history of multiple frames of the video, generating, by the ANN, a prediction of the current frame of the video based on the reference frame and the current frame of the video; 16. The apparatus, configured to generate, by an ANN, a reconstruction of a current frame based on a prediction of a current frame of the video and a residual calculated based on a difference between the current frame of the video and the prediction of the current frame of the video. 19. The apparatus of clause 18, wherein the history includes a plurality of previously decoded frames. 20. The apparatus of clause 18 or 19, wherein at least one processor is further configured to encode a difference between a current frame of video and a prediction of the current frame of video, and add the encoding to the prediction of the current frame.
[0190] In one aspect, the receiving means, extracting means, determining means, transmitting means, decoding means, means for generating a prediction by an ANN, means for generating a prediction by an ANN, means for generating a reference frame by an ANN, and / or means for generating a reconstruction by an ANN may be the CPU 102, a program memory associated with the CPU 102, the dedicated memory block 118, the fully connected layer 362, the NPU 428 / , and / or the routing connection processing unit 216 configured to perform the functions described. In another configuration, the above-mentioned means may be any module or any device configured to perform the functions recited by the above-mentioned means.
[0191]
[0084] The various operations of the methods described above may be implemented by any suitable means capable of performing the corresponding functions. The means may include various hardware and / or software component(s) and / or module(s), including but not limited to circuits, application specific integrated circuits (ASICs), or processors. In general, where there are operations illustrated in a figure, those operations may have corresponding counterpart means-plus-function components that are similarly numbered.
[0192]
[0085] As used herein, the term "determining" encompasses a wide variety of actions. For example, "determining" may include calculating, computing, processing, deriving, investigating, looking up (e.g., looking up in a table, database, or another data structure), ascertaining, and the like. In addition, "determining" may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), and the like. Further, "determining" may include resolving, selecting, choosing, establishing, and the like.
[0193]
[0086] As used herein, a phrase referring to "at least one of" a list of items refers to any combination of those items, including single members. As an example, "at least one of a, b, or c" is intended to include a, b, c, ab, ac, bc, and abc.
[0194]
[0087] The various example logic blocks, modules and circuits described in connection with this disclosure may be implemented or performed using a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the described functions. A general purpose processor may be a microprocessor, but alternatively, the processor may be any commercially available processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.
[0195]
[0088] The steps of a method or algorithm described in relation to the present disclosure may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in any form of storage medium known in the art. Some examples of storage media that may be used include random access memory (RAM), read only memory (ROM), flash memory, erasable programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), registers, hard disks, removable disks, CD-ROMs, etc. A software module may comprise a single instruction or many instructions, and may be distributed across several different code segments, among different programs, across multiple storage media. The storage medium may be coupled to the processor such that the processor can read information from and write information to the storage medium. Alternatively, the storage medium may be integral to the processor.
[0196]
[0089] The methods disclosed herein include one or more steps or actions for achieving the described method. The steps and / or actions of those methods may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims.
[0197]
[0090] The described functions may be implemented in hardware, software, firmware, or any combination thereof. When implemented in hardware, an exemplary hardware configuration may comprise a processing system in a device. The processing system may be implemented using a bus architecture. The bus may include any number of interconnected buses and bridges depending on the particular application of the processing system and the overall design constraints. The bus may link various circuits together, including the processor, the machine-readable medium, and the bus interface. The bus interface may be used to connect, among other things, a network adapter to the processing system via the bus. The network adapter may be used to implement signal processing functions. In some aspects, a user interface (e.g., a keypad, a display, a mouse, a joystick, etc.) may also be connected to the bus. The bus may also link various other circuits, such as timing sources, peripherals, voltage regulators, power management circuits, etc., which are well known in the art and therefore will not be described further.
[0198]
[0091] The processor may be responsible for managing the bus and general processing including executing software stored on the machine-readable medium. The processor may be implemented using one or more general-purpose processors and / or special-purpose processors. Examples include microprocessors, microcontrollers, DSP processors, and other circuitry capable of executing software. Software shall be broadly construed to mean instructions, data, or any combination thereof, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise. The machine-readable medium may include, by way of example, a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a register, a magnetic disk, an optical disk, a hard drive, or any other suitable storage medium, or any combination thereof. The machine-readable medium may be embodied in a computer program product. The computer program product may include packaging materials.
[0199]
[0092] In a hardware implementation, the machine-readable medium may be part of a processing system separate from the processor. However, as one skilled in the art will readily appreciate, the machine-readable medium or any part thereof may be external to the processing system. By way of example, the machine-readable medium may include a transmission line, a carrier wave modulated by data, and / or a computer product separate from the device, all of which may be accessed by the processor through a bus interface. Alternatively or additionally, the machine-readable medium or any part thereof may be integrated into the processor, as may be a cache and / or a general-purpose register file. Although the various components discussed may be described as having a particular location, such as a local component, these components may also be configured in various ways, such as some components configured as part of a distributed computing system.
[0200]
[0093] The processing system may be configured as a general-purpose processing system having one or more microprocessors providing processor functions and external memory providing at least a portion of the machine-readable medium, all linked together with other support circuits through an external bus architecture. Alternatively, the processing system may include one or more neuromorphic processors for implementing the described neuron model and model of the neural system. As another alternative, the processing system may be implemented using an application specific integrated circuit (ASIC) having a processor, bus interface, user interface, support circuits, and at least a portion of the machine-readable medium integrated into a single chip, or using one or more field programmable gate arrays (FPGAs), programmable logic devices (PLDs), controllers, state machines, gate logic, discrete hardware components, or any other suitable circuitry, or any combination of circuits capable of performing the various functions described throughout this disclosure. Those skilled in the art will recognize how to best implement the described functions for the processing system depending on the particular application and the overall design constraints imposed on the entire system.
[0201]
[0094] The machine-readable medium may comprise several software modules. The software modules include instructions that, when executed by the processor, cause the processing system to perform various functions. The software modules may include a transmitting module and a receiving module. Each software module may reside in a single storage device or may be distributed across multiple storage devices. As an example, the software modules may be loaded into RAM from a hard drive when a triggering event occurs. During execution of the software modules, the processor may load some of the instructions into a cache to increase access speed. One or more cache lines may then be loaded into a general purpose register file for execution by the processor. When referring to functions of a software module below, it will be understood that such functions are implemented by the processor upon executing instructions from that software module. Furthermore, it will be understood that aspects of the present disclosure provide for improved functionality of a processor, computer, machine, or other system implementing such aspects.
[0202]
[0095] If implemented in software, the functions may be stored on or transmitted via a computer-readable medium as one or more instructions or code. Computer-readable media includes both computer storage media and communication media, including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. By way of example, and not limitation, such computer-readable media may comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. In addition, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared (IR), radio, and microwave, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. As used herein, disk and disc include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc, where disks typically reproduce data magnetically and discs reproduce data optically using lasers. Thus, in some aspects computer readable medium may comprise non-transitory computer readable medium (e.g., tangible media). In addition, in other aspects computer readable media may comprise transitory computer readable media (eg, a signal). Combinations of the above should also be included within the scope of computer readable media.
[0203]
[0096] Accordingly, some aspects may include a computer program product for performing the operations presented herein. For example, such a computer program product may comprise a computer-readable medium having stored thereon (and / or encoded thereon) instructions executable by one or more processors to perform the described operations. In some aspects, the computer program product may include packaging materials.
[0204]
[0097] Furthermore, it should be understood that modules and / or other suitable means for performing the described methods and techniques may be downloaded and / or otherwise obtained by a user terminal and / or base station, where applicable. For example, such devices may be coupled to a server to facilitate the transfer of means for performing the described methods. Alternatively, the various methods described may be provided via a storage means, such that the user terminal and / or base station may obtain the various methods upon coupling or providing the storage means (e.g., RAM, ROM, physical storage medium such as a compact disk (CD) or floppy disk, etc.) to the device. Furthermore, any other suitable technique for providing the described methods and techniques to the device may be utilized.
[0205]
[0098] It should be understood that the claims are not limited to the precise configuration and components illustrated above. Various modifications, changes and variations may be made in the configuration, operation and details of the methods and apparatus described above without departing from the scope of the claims.
Claims
1. 1. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving a video by said ANN; extracting, by the ANN, a first set of features of a current frame of the video and a second set of features of a reference frame of the video; determining, by the ANN, an estimate of a correlation feature between the first set of features of the current frame and the second set of features of the reference frame; transmitting said estimate of the correlation characteristic to a receiver; 13. A processor-implemented method comprising:
2. The processor-implemented method of claim 1 , further comprising determining the estimate of a correlation feature via a full convolution operator of the ANN.
3. The processor-implemented method of claim 1 , wherein the reference frame comprises a reconstruction of a previous frame of the video.
4. The processor-implemented method of claim 1 , wherein the reference frame is predicted based on a reconstruction of a number of previous frames of the video.
5. 1. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving, by the ANN, a reference frame of a video; receiving, by the ANN, an estimate of a correlation feature between a current frame of the video and the reference frame of the video; extracting, by the ANN, a first set of features of the reference frames of the video; Decoding the estimate of correlation features by the ANN; and generating, by the ANN, a prediction of the current frame of the video based on the decoded estimate of correlation features and the first set of features of the reference frames of the video; 13. A processor-implemented method comprising:
6. calculating a residual based on a difference between the current frame of the video and the prediction of the current frame of the video; generating a reconstruction of the current frame of the video based on the prediction of the current frame and the residual; The processor-implemented method of claim 5 further comprising:
7. The processor-implemented method of claim 5 , wherein the reference frame is predicted based on a reconstruction of a number of previous frames of the video.
8. 1. A processor-implemented method for video compression using an artificial neural network (ANN), comprising: receiving, by said ANN, a history of a plurality of frames of a video; generating a reference frame from the history of a plurality of frames of the video using the ANN; generating, by the ANN, a prediction of the current frame of the video based on the reference frame and the current frame of the video; generating, by the ANN, a reconstruction of the current frame based on the prediction of the current frame of the video and a residual calculated based on a difference between the current frame of the video and the prediction of the current frame of the video; 13. A processor-implemented method comprising:
9. The processor-implemented method of claim 8 , wherein the history comprises a plurality of previously decoded frames.
10. 10. The processor-implemented method of claim 8, further comprising: encoding the difference between the current frame of the video and the prediction of the current frame of the video; and adding the encoding to the prediction of the current frame.
11. 1. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: receiving a video by said ANN; extracting, by the ANN, a first set of features for a current frame of the video and a second set of features for a reference frame of the video; determining, by the ANN, an estimate of correlation features between the first set of features of the current frame and the second set of features of the reference frame; The apparatus being configured to transmit the estimate of the correlation feature to a receiver.
12. The at least one processor: The apparatus of claim 11 , further configured to determine the estimate of a correlation feature via a full convolution operator of the ANN.
13. The apparatus of claim 11 , wherein the reference frame comprises a reconstruction of a previous frame of the video.
14. The apparatus of claim 11 , wherein the at least one processor is further configured to predict the reference frame based on a reconstruction of a number of previous frames of the video.
15. 1. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: receiving a reference frame of the video; receiving an estimate of a correlation feature between a current frame of the video and the reference frame of the video; Extracting a first set of features of the reference frame of the video; Decoding said estimate of correlation features; An apparatus configured to generate a prediction of the current frame of the video based on the decoded estimate of correlation features and the first set of features of the reference frames of the video.
16. The at least one processor: calculating a residual based on a difference between the current frame of the video and the prediction of the current frame of the video; The apparatus of claim 15 , further configured to generate a reconstruction of the current frame of the video based on the prediction and the residual of the current frame.
17. The apparatus of claim 15 , wherein the at least one processor is further configured to predict the reference frame based on a reconstruction of a number of previous frames of the video.
18. 1. An apparatus for video compression using an artificial neural network (ANN), comprising: Memory, at least one processor coupled to the memory, the at least one processor comprising: receiving, by said ANN, a history of a number of frames of a video; generating a reference frame from the history of frames of the video using the ANN; generating, by the ANN, a prediction of the current frame of the video based on the reference frame and the current frame of the video; and generating, by the ANN, a reconstruction of the current frame based on the prediction of the current frame of the video and a residual calculated based on a difference between the current frame of the video and the prediction of the current frame of the video.
19. The apparatus of claim 18 , wherein the history comprises a plurality of previously decoded frames.
20. 20. The apparatus of claim 18, wherein the at least one processor is further configured to encode the difference between the current frame of the video and the prediction of the current frame of the video and add the encoding to the prediction of the current frame.
Citation Information
Patent Citations
Deformable Convolutional Predictive Frame Generation for Video Coding
JP2023502535A
Temporal Structure Based Conditional Convolutional Neural Networks for Video Compression
JP2024510433A
Method and apparatus for video encoding and video decoding based on neural network
US20190246102A1
Systems and methods for generating a latent space residual
US20210067808A1
