Feature detection model for autonomous and semi-autonomous systems and applications

Abstract

The invention relates to a feature detection model for autonomous and semi-autonomous systems and applications. In various examples, feature detection models for autonomous and / or semi-autonomous systems and applications are described herein. The systems and methods described herein may use one or more trained machine learning models to automatically generate representations of traffic features corresponding to a map, such as road signs and / or road edges. For example, a model may take as input an image representing at least a portion of a map that includes one or more traffic features and one or more indications of one or more points associated with the traffic features represented by the image. Based at least on processing these inputs, the model may generate and / or output data representing additional points and / or heat maps associated with the traffic feature, which represent one or more lines representing the traffic feature. The output data may then be used to determine a representation of the traffic feature for annotating the map.

Description

Background Technology

[0001] For vehicles (such as autonomous vehicles, semi-autonomous vehicles, robots, etc.) to operate safely in their environment, they must be able to effectively perform vehicle maneuvers—such as lane keeping, lane changing, lane splitting, turning, stopping and starting at intersections, crosswalks, etc., and / or other vehicle or machine maneuvers. For example, for vehicles to navigate on paved streets (such as city streets, side streets, community streets, etc.) and highways (such as multi-lane roads), they need to navigate within one or more divisions or demarcations (such as lanes, intersections, crosswalks, boundaries, etc.), which are typically marked using traffic features—such as road markings including arrows, text, graphics, and / or other types of content. Therefore, it is crucial for vehicles to be able to detect traffic features in their environment so that they can determine how to drive based on the rules associated with those features.

[0002] To detect traffic features, vehicles can at least partially use a map corresponding to the environment in which they are driving. For example, the map can be annotated to indicate the location of important traffic features that the vehicle needs to identify while driving, such as road edges, road markings, traffic signs, etc. Some common methods for annotating such maps involve the user viewing different sections of the map to manually enter labels for the traffic features. For example, for a road marking of a given length, the user can manually indicate the location of the road marking by selecting multiple points (e.g., hundreds and / or thousands) along the road marking. However, having users manually indicate the location of traffic features represented by the map can be time-consuming, prone to user errors, and / or require significant computational resources (depending on the user's device).

[0003] Therefore, more specifically, for road signs, other conventional methods can use curve fitting functions to connect existing road signs that have already been annotated on the map. For example, if a user has annotated the first part of a road sign and the second part of a separate road sign, these conventional methods will simply use a curve fitting function to attach these two parts of the road sign together. However, while these conventional methods may be accurate for straight road signs by simply connecting existing road signs using a curve fitting function, they are inaccurate for road signs that include one or more curves. Furthermore, because these conventional methods operate on the entire map, when the map is segmented into sub-parts (e.g., images), such as those used to provide the map to vehicles for navigation, the annotations of the generated road signs may be misaligned. Summary of the Invention

[0004] Embodiments of this disclosure relate to feature detection models for autonomous and / or semi-autonomous systems and applications. The systems and methods described herein can use one or more trained machine learning models (one or more models) to automatically generate representations of traffic features corresponding to a map, such as road signs and / or road edges. For example, one or more models can take as input an image representing at least a portion of a map including one or more traffic features, and one or more indicators of one or more points (e.g., one or more cues) associated with the one or more traffic features represented by the image. Based at least on processing these inputs, one or more models can generate and / or output data representing additional points and / or heatmaps (which represent one or more lines corresponding to one or more traffic features) associated with the one or more traffic features. In some examples, one or more models and / or another post-processing component can then use the output (e.g., line representations of road signs and / or road edges) to determine one or more final representations of the one or more traffic features, which can then be used to annotate the map.

[0005] Compared to conventional systems, in some embodiments, the system of this disclosure is capable of automatically determining the location of traffic features represented by a map using prompts and / or input images. Thus, the system of this disclosure does not require the user to manually input all points of the traffic features (e.g., hundreds and / or thousands of points) when annotating the map. Furthermore, as described in more detail herein, one or more models can be trained to determine multiple points (e.g., up to one hundred or more points) associated with the traffic features, and these points are then used to determine the final representation of the traffic features within the map (e.g., a line representation). Therefore, compared to previous systems that only use line fitting to connect lines already labeled on the map, the system of this disclosure can generate more accurate annotations of traffic features. Attached Figure Description

[0006] The present system and method for feature detection models used in autonomous and / or semi-autonomous systems and applications are described in detail below with reference to the accompanying drawings, wherein:

[0007] Figure 1A An example data flow diagram illustrates a process for detecting features and / or annotating maps using one or more machine learning models according to some embodiments of this disclosure;

[0008] Figure 1B Examples of the architecture of one or more machine learning models for detecting features according to some embodiments of the present disclosure are shown;

[0009] Figure 1C Examples of cross-attention layers that can be used by one or more machine learning models according to some embodiments of this disclosure are shown;

[0010] Figure 1D Examples of one or more machine learning models are shown, illustrating the information associated with points of features in relation to the output of some embodiments of this disclosure;

[0011] Figure 2A-2B Examples of segmenting maps to generate images representing features within an environment, according to some embodiments of the present disclosure, are shown.

[0012] Figure 3 Examples of determining the location of input points associated with features represented by an image, according to some embodiments of the present disclosure, are shown;

[0013] Figure 4 Examples of determining points associated with features represented by an image, according to some embodiments of this disclosure, are shown;

[0014] Figure 5 Examples of heatmaps indicating the location of features represented by images according to some embodiments of the present disclosure are shown;

[0015] Figure 6 Examples of annotated maps according to some embodiments of this disclosure are shown to include information associated with features;

[0016] Figure 7 The diagram illustrates a data flow diagram of a process for training one or more machine learning models to detect features associated with an image and / or map, according to some embodiments of the present disclosure.

[0017] Figure 8 Examples of one or more systems that can be configured to perform at least a portion of the processes described herein, according to some embodiments of the present disclosure, are shown.

[0018] Figure 9-10 A flowchart is shown according to some embodiments of the present disclosure, illustrating a method for detecting features associated with an image;

[0019] Figure 11A These are illustrations of example autonomous vehicles according to some embodiments of the present disclosure;

[0020] Figure 11B According to some embodiments of this disclosure Figure 11A Examples of camera positions and fields of view for autonomous vehicles;

[0021] Figure 11C According to some embodiments of this disclosure Figure 11A A block diagram of an example system architecture for an example autonomous vehicle;

[0022] Figure 11DThis is based on some embodiments of the present disclosure for use in cloud-based servers and Figure 11A A system diagram illustrating communication between autonomous vehicles;

[0023] Figure 12 This is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure; and

[0024] Figure 13 This is a block diagram of an example data center suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0025] Systems and methods related to feature detection models for autonomous and / or semi-autonomous systems and applications are disclosed. Although this disclosure may relate to an example autonomous or semi-autonomous vehicle or machine 1100 (which may be alternatively referred to herein as "vehicle 1100", "self-vehicle 1100", "self-machine 1100" or "machine 1100"), examples are related to... Figure 11A-11D The description herein is intended to be restrictive. For example, the systems and methods described herein may be used, 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, driver- and driverless robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, spacecraft, ships, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, submarines, drones, and / or other vehicle types. Furthermore, although this disclosure may relate to the detection of traffic features and / or the annotation of maps with labels associated with traffic features, this is not intended to be restrictive, and the systems and methods described herein can be used in augmented reality, virtual reality, mixed reality, robotics, security and supervision, autonomous or semi-autonomous machine applications, and / or any other technical field that may utilize object detection or feature detection and / or map creation.

[0026] For example, the system can generate, receive, retrieve, acquire, and / or store map data representing maps associated with the environment (e.g., navigation maps, standard definition (SD) maps, high definition (HD) maps, occupancy maps, or grids and / or any other type of map). In some examples, the map can be associated with different layers corresponding to various types of sensor modalities and / or various types of features. For example, the map can be associated with one or more image layers, one or more LiDAR layers, one or more RADAR layers, one or more waiting condition layers, one or more road sign layers, and / or any other type of layer. Furthermore, the map can represent information associated with features located within the environment, including but not limited to the location of the features, the orientation of the features, the classification of the features, and / or any other information. As described herein, features can include traffic features (e.g., road edges, road signs, lane edges, lane markings, traffic poles, traffic signs, traffic lights, and / or the like) and / or any other type of features and / or objects that may be located within the environment.

[0027] The system can then perform various techniques to generate annotations for at least some features represented by the map. For example, the system can segment the map into multiple parts representing specific areas of the environment. As described herein, in some examples, images representing different areas of the environment (e.g., top-down images (e.g., bird's-eye view) and / or intensity images) can be used to represent different parts of the map. For example, LiDAR data associated with the map (e.g., LiDAR layers) and / or point cloud data associated with the LiDAR data can represent the intensity of points within the environment. For example, a LiDAR sensor used to generate the LiDAR data can measure the intensity of points as light returns to the LiDAR sensor. Intensity can be represented numerically, such as a number between 0 and 256 (and / or any other range), where the number varies depending on the composition of the surface reflecting the light (e.g., color, texture, material, etc.). For example, a low value can indicate low reflectivity, while a high value can indicate high reflectivity. In some examples, intensity can depend on other factors, such as the angle of arrival, the extent of the point, etc.

[0028] Therefore, the system can process LiDAR data and / or point cloud data to generate images associated with a map (e.g., top-down intensity images). Since the intensity of a point can vary based on one or more factors, such as the color of the surface associated with the point (e.g., the color of the surface reflecting light), the image can indicate the structure of various features located within the environment. For example, if the features include road markings on a road surface within the environment, where the road markings are drawn using a specific color (e.g., white, yellow, etc.), the image of the environment can depict points associated with the road markings as a different color compared to points associated with other features (e.g., the road surface itself).

[0029] The system can then use the image to detect features located in the environment. For example, with an image, the system can determine one or more points (one or more cue points) associated with features represented by the image (e.g., lane markings). In some examples, the system can use input data representing one or more inputs from one or more users, indicating the location of the cue points within the image. Additionally or alternatively, in some examples, the system can process the image and / or map to automatically determine the location of the cue points within the image. For example, a portion of the features represented by the map may have been annotated, for example, by using one or more processes described herein. Thus, the system can use the annotations to determine the location of the cue points within the image.

[0030] The system can then use one or more machine learning models (e.g., one or more models) to process the input data associated with the image and / or cue points. These machine learning models are trained to determine information associated with features, such as points (e.g., output points) and / or representations of features. For example, the model may include one or more encoders (one or more image encoders) configured to generate one or more image embeddings associated with the image. The model may also include one or more supplemental encoders (one or more cue encoders) configured to generate one or more tokens associated with the cue points. In some examples, the cue encoder may be configured to generate a corresponding token associated with each cue point. In some examples, the model may append tokens to one or more learnable tokens to generate one or more input tokens, which will be described in more detail herein. However, as described in more detail herein, in some examples, the system may not receive any cue points associated with the image. In such examples, the input tokens may consist only of learnable tokens.

[0031] The model may then include one or more cross-attention components configured to process image embeddings and input tokens. For example, in some examples, a cross-attention component may include one or more cross-attention layers (e.g., one or more bidirectional cross-attention layers), one or more token-image converters, and / or any other type of processing component configured to perform one or more processes described herein. Based at least on this processing, the cross-attention component may generate and / or output at least one token (e.g., output token) associated with the location of the cue point and at least one token (e.g., image token) associated with the image. In some examples, the output token may be associated with a vector including a specific dimension.

[0032] The model can then process the output labels and / or image labels using one or more decoders configured to determine output points associated with the features. In some examples, the decoders can include any type of decoder, such as an autoregressive transformer decoder. Furthermore, in some examples, the decoders can be configured to use a sequence to determine a threshold number of output points (e.g., 100). For example, the decoder can use the output labels to determine a first output point, use a second output label associated with the first output point (and / or any other preceding output label) to determine a second output point, use a third output label associated with the second output point (and / or any other preceding output label) to determine a third output point, use a fourth output label associated with the third output point (and / or any other preceding output label) to determine a fourth output point, and so on, until the threshold number of points is reached.

[0033] In some examples, the decoder uses a matrix (which can represent one or more learnable tags) to determine additional output tags, including dimensions associated with the image size. For example, if the image size is 1024*512, the matrix dimensions could include 1024 units in the x-direction and 512 units in the y-direction. To determine the output tags, the decoder can use the matrix and a previous output point. For example, to determine the second output tag in the example above, the decoder can use the matrix and the first output point to determine a vector associated with the first output point, where this vector is associated with the second output tag. Furthermore, to determine the third output tag, the decoder can use the matrix and the second output point to determine a vector associated with the second output point, where this vector is associated with the third output tag.

[0034] Furthermore, in some examples, the decoder can be configured to determine the classification associated with the output point. As described herein, the classification can include, but is not limited to, valid points (e.g., points associated with a feature), invalid points (e.g., points not associated with a feature, which may be discarded), joint points (e.g., points associated with multiple features), and / or any other classification.

[0035] The model can also use one or more additional layers to process image labels and / or output labels, these additional layers being configured to generate heatmaps indicating the locations of features represented by the image. In some examples, the additional layers can include any type of layer associated with the machine learning model, such as one or more convolutional layers. The system and / or model can then use the output points and / or heatmaps to determine the final representation associated with the features. For example, since the output points can be generated using a sequence, the system and / or model can generate the final representation by connecting the individual points (e.g., valid points and / or joint points) according to the order in which the points were generated. Furthermore, the system and / or model can use heatmaps, such as to improve the connectivity of points. By performing such a process, the system can generate representations of features within the environment, such as line representations of road edges, road markings, lane edges, and / or lane markings.

[0036] In some examples, the system may perform a similar process to generate one or more additional representations of one or more additional features represented by an image and / or one or more additional images. Furthermore, in some examples, the system may perform one or more processes to determine the classification associated with the feature. For example, the system may process a map, images of features from the map (e.g., intensity images, color images, etc.), and / or any other information associated with the feature to determine the final classification. For example, if the feature includes road signs, the system may use a model (e.g., one or more additional layers of the model) and / or one or more additional machine learning models trained to process the image to determine whether the road sign includes solid line road signs, dashed line road signs, double line road signs, center road signs, two-way road signs, overtaking road signs, arrows, stop lines, pedestrian crossings, and / or any other type of road sign. Furthermore, in some examples, the system may then use these determinations to annotate the map, for example, by labeling the map to indicate the location and / or classification associated with the feature.

[0037] Therefore, by executing one or more of the processes described herein, the system can determine the final representation associated with a feature with little or no input from one or more users. For example, when the system uses only learnable labels as input labels for the model, it can automatically determine the final representation without input from one or more users. For instance, if the features include road signs in the environment, the system can use learnable labels and a model to automatically determine the line representation associated with the road signs. This offers many improvements over conventional systems, which also require users to select hundreds and / or thousands of points to generate such line representations and / or use only curve fitting to generate them.

[0038] In some examples, the system can train a model to perform one or more processes described herein, such as determining information associated with features. For example, the model can be trained using training input data and corresponding ground truth data, such as training images (e.g., top-down images and / or intensity images) representing the locations of features and / or cue points associated with those features, and corresponding ground truth data such as output points (e.g., actual points) associated with features, heatmaps indicating the locations of features, and / or classification data associated with the output points. One or more training engines can then be configured to determine one or more losses using the outputs from the model generated based on the processing of the training input data and ground truth data. For example, the training engines can determine one or more losses at least based on comparing the outputs with ground truth data using one or more loss functions. The training engines can also be configured to update one or more parameters and / or weights associated with the model using one or more losses. The training of the model is described in more detail.

[0039] While the examples in this paper describe processing images associated with a map to determine information associated with features represented by the map, similar processes can be used in other examples with other techniques. For instance, a machine navigating an environment can generate images using one or more sensors, such as one or more red-green-blue images, one or more intensity images, one or more top-down images, and / or any other type of image. The machine can then process the images using one or more processes described herein to determine information associated with features located in the environment. Furthermore, the machine can perform one or more operations based at least on the information associated with the features. For example, when the features include road signs, the machine can determine how to navigate in the environment based at least on the location of the road signs in the environment.

[0040] In some examples, the machine learning models described herein (e.g., deep neural networks, language models, LLMs, VLMs, multimodal language models, perceptual models, tracking models, fusion models, transformer models, diffusion models, encoder-only models, decoder-only models, encoder-decoder models, neural rendering field (NERF) models, etc.) can be packaged as microservices (such as inference microservices (e.g., NVIDIA NIM)), which may include containers (e.g., operating system (OS) level virtualization packages) that may include an application programming interface (API) layer, a server layer, a runtime layer, and / or a model "engine". For example, an inference microservice may include the container itself and the model (e.g., weights and biases). In some cases, such as when the machine learning model is small enough (e.g., has a sufficiently small number of parameters), the model may be included within the container itself. In other examples (e.g., when the model is large), the model may be hosted / stored in the cloud (e.g., in a data center) and / or may be hosted on-premise and / or at the edge (e.g., on a local server or computing device, but outside the container). In such embodiments, the model may be accessed via one or more APIs (e.g., REST APIs). Therefore, in some embodiments, the machine learning models described herein can be deployed as inference microservices to accelerate model deployment on any cloud, data center, or edge computing system while ensuring data security. For example, an inference microservice may include one or more APIs, pre-configured containers for simplified deployment, an optimized inference engine (e.g., execution software built using standardized AI model deployments, such as NVIDIA's Triton Inference Server) and / or one or more APIs for high-performance deep learning inference, which may include inference runtime and model optimizations that can provide low latency and high throughput for production applications (e.g., NVIDIA's TensorRT), and / or enterprise management data for telemetry (e.g., including identity, metrics, health checks, and / or monitoring). The machine learning models described herein may be included as part of a microservice along with an acceleration infrastructure capable of deployment using a single command and / or orchestration and autoscaling (e.g., reaching data center scale on a single device) using a container orchestration system on the acceleration infrastructure. Therefore, an inference microservice may include a machine learning model (e.g., optimized for high-performance inference), inference runtime software for executing the machine learning model and providing output / response to inputs (e.g., user queries, prompts, etc.), and enterprise management software for providing health checks, identity and / or other monitoring. In some embodiments, the inference microservice may include software for performing in-situ replacements and / or updates to the machine learning model.When replacing or updating, the software performing the replacement / update can maintain the user configurations of the inference runtime software and the enterprise management software.

[0041] Furthermore, in some embodiments, the systems and methods described herein can be performed in a simulated environment (e.g., NVIDIA's DriveSIM) using simulated data (e.g., simulated sensor data from simulated sensors of a virtual or simulated machine). For example, simulated sensor data and / or map data (simulated or real) can be used to perform various operations in the simulated environment, such as determining the location of features in the simulated environment. These simulated operations can be used to test the performance of underlying algorithms, systems, and / or processes before deploying them to the real world. In some cases, simulations can be used to generate synthetic training data (e.g., training data including landmarks, features, objects, etc.), which can then be processed (in addition to or as a substitute for real-world data) to perform feature detection, update virtual or simulated maps, and / or perform other operations.

[0042] In any example, such as when a simulation environment is used for testing, validation, training, etc., one or more optical transport algorithms (e.g., ray tracing and / or path tracing algorithms) may be used to render or otherwise generate the simulation environment and / or associated training data. In some embodiments, simulation environments and / or one or more of their objects, features, or components may be generated or managed within a 3D content collaboration platform (e.g., NVIDIA's OMNIVERSE) for industrial digitization, generative physics AI, and / or other use cases, applications, or services. For example, the content collaboration platform or system may include systems for using or developing generic scene descriptors (USD) (e.g., OpenUSD) data to manage objects, features, scenes, etc., within simulation environments, digital environments, etc. The platform may include realistic physics simulations, such as using NVIDIA's PhysX SDK, to simulate real physics and physical interactions with the simulation hosted by the platform. This platform can integrate OpenUSD with ray tracing / path tracing / light transport simulations (such as NVIDIA's RTX rendering technology) into software tools and simulation workflows to build, train, deploy, or test AI systems—such as systems for testing, validating, training (e.g., machine learning models, neural networks, etc.) and / or other tasks related to automobiles, robots, machines, or other applications.

[0043] The systems and methods described herein can be used, 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, driver- and driverless robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, spacecraft, ships, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, submarines, drones, and / or other vehicle types. Furthermore, the systems and methods described herein can be used for a variety of purposes, including, but not limited to, machine control, machine motion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, safety and supervision, 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, optical transport simulation (e.g., ray tracing, path tracing, etc.), collaborative content creation of 3D assets, cloud computing, and / or any other suitable application.

[0044] The disclosed embodiments can 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, aviation 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 implementing large language models (LLMs), systems implementing one or more multimodal language models, systems using or deploying one or more inference microservices, systems containing or deploying one or more machine learning models and operating system-level virtualization packages (e.g., containers) in services or microservices, systems containing one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implemented at least partially in a data center, systems for performing conversational AI operations, systems for performing optical transmission simulations, systems for performing collaborative content creation of 3D assets, systems for performing generative AI operations, systems implemented at least partially using cloud computing resources, and / or other types of systems.

[0045] refer to Figure 1A , Figure 1AExample data flow diagrams are shown for a process 100 of detecting features and / or annotating maps using one or more machine learning models 102 according to some embodiments of this disclosure. It should be understood that such and other arrangements described herein are presented by way of example only. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groups, etc.) may be used in addition to or in place of the arrangements and elements shown, and some elements may be omitted entirely. Furthermore, many of the elements described herein are functional entities that can be implemented as discrete or distributed components or in combination with other components, and can be implemented in any suitable combination and location. The various functions performed by the entities described herein can be performed by hardware, firmware, and / or software. For example, various functions can be performed by a processor executing instructions stored in memory. In some embodiments, the systems, methods, and processes described herein can use... Figure 11A-11D Example autonomous vehicles 1100 Figure 12 Example computing device 1200 and / or Figure 13 The example data center 1300 uses similar components, features, and / or functions to perform this.

[0046] For example, process 100 may include: one or more segmentation components 104 receiving map data 106 representing one or more maps. As described herein, maps may include any type of map associated with the environment, such as navigation maps, SD maps, HD maps, and / or any other type of map. In some examples, maps may be associated with different layers corresponding to various types of sensor modalities and / or various types of features. For example, maps may be associated with one or more image layers, one or more LiDAR layers, one or more RADAR layers, one or more waiting condition layers, one or more lane marking layers, and / or any other type of layer. Furthermore, maps may represent information associated with features located within the environment, including but not limited to the location of features, the orientation of features, the classification of features, and / or any other information. As described herein, features may include traffic features (e.g., road edges, road markings, lane edges, lane markings (e.g., Class I road markings), traffic poles, traffic signs, traffic lights, and / or the like) and / or any other type of features and / or objects that may be located within the environment.

[0047] Then, process 100 may include: segmentation component 104 segmenting the map into multiple parts representing specific regions of the environment. As described herein, in some examples, images representing different regions of the environment (e.g., top-down images (e.g., bird's-eye view) and / or intensity images) may be used to represent the parts of the map. For example, LiDAR data associated with the map (e.g., LiDAR layers) and / or point cloud data associated with the LiDAR data may represent the intensity of points within the environment. For example, a LiDAR sensor used to generate the LiDAR data may measure the intensity of points as light returns to the LiDAR sensor. Intensity may be represented numerically, such as a number between 0 and 256 (and / or any other range), where the number varies based on the composition of the surface reflecting the light (e.g., color, texture, material, etc.). For example, a low value may indicate low reflectivity, while a high value may indicate high reflectivity. In some examples, intensity may depend on other factors, such as the angle of arrival, the extent of the point, etc.

[0048] Therefore, segmentation component 104 can process LiDAR data and / or point cloud data to generate an image associated with a map (e.g., a top-down intensity image). Since the intensity of a point can vary based on one or more factors, such as the color of the surface associated with the point (e.g., the color of the surface reflecting light), the image can indicate the structure of various features located within the environment. For example, if the features include road markings on a road within the environment, where the road markings are drawn using a specific color (e.g., white, yellow, etc.), the image of the environment can depict points associated with road markings as a different color compared to points associated with other features (e.g., the road surface itself).

[0049] For example, Figure 2A-2B Examples of segmenting maps according to some embodiments of this disclosure to generate images representing features within an environment are shown. Figure 2A As shown in the example, segmentation component 104 may first receive a map 202 of the environment, wherein the environment includes at least features 204(1)-(8) (also referred to singularly as “feature 204” or plurally as “multiple features 204”). Although Figure 2A The example shows feature 204 as including road and / or lane markings, but in other examples, map 202 could represent any other type of feature located within the environment. Segmentation component 104 can then segment map 202 into multiple parts 206(1)-(4) (also referred to singularly as “part 206” or plurally as “multiple parts 206”). Although Figure 2AThe example shows map 202 being divided into four non-overlapping sections 206, but in other examples, the segmentation component 104 may divide map 202 into any number of sections and / or one or more sections may overlap with one or more other sections.

[0050] Next, as Figure 2B As shown in the example, segmentation component 104 can generate at least an image 208 associated with the first portion 206(1) of map 202. Figure 2B In the example, image 208 may include a top-down intensity image generated by segmentation component 104 using LiDAR data and / or point cloud data associated with map 202 and / or the environment. For example, pixels in image 208 may indicate the intensity of points within the environment as measured by one or more LiDAR sensors when generating the LiDAR data. Thus, pixels associated with feature 204 may be associated with points including a first intensity, while pixels associated with other features (e.g., road surfaces, surrounding surfaces, etc.) may be associated with points including a lower second intensity. Therefore, feature 204 appears brighter within image 208 compared to the surrounding environment.

[0051] Although Figure 2B The example shows the generation of a top-down intensity image 208, but in other examples, segmentation component 104 can generate any other type of image. For example, segmentation component 104 can generate a red-green-blue (RGB) image associated with the first portion 206(1) of map 202, for example, by using image data associated with map 202 and / or the environment. Furthermore, although Figure 2A-2B The example shows the generation of an image 208 associated with the first part 206(1) of map 202, but in other examples, segmentation component 104 may generate one or more additional images associated with one or more of the multiple parts 206(2)-(4) of map 202.

[0052] Return to reference Figure 1AFor example, process 100 may include: one or more user devices 108 receiving image data 110 representing an image generated by segmentation component 104. Based at least on the received image data 110, user device 108 may present the image to one or more users, who can then view the image and provide input associated with features represented by the image. For example, as described herein, a user may indicate one or more locations of one or more points (referred to in some examples as “cue points”) located on corresponding features represented by the image. For example, if the features include road signs, the user may indicate the location of a point located on a road sign. Based at least on these inputs, process 100 may include: user device 108 generating and / or outputting point data 112 representing the points input by the user.

[0053] For example, Figure 3 Examples of determining the positions of cue points 302(1)-(3) (also referred to singularly as “cue point 302” or plurally as “a plurality of cue points 302”) associated with feature 204 represented by image 208 according to some embodiments of the present disclosure are shown. As shown, the cue points 302 include at least a first cue point 302(1) located on a first feature 204(1), a second cue point 302(2) located on a second feature 204(2), and a third cue point 302(3) located on a third feature 204(3). Although Figure 3 The example shows only one cue point 302 for each feature 204, but in other examples, any number of cue points can be indicated for each feature 204.

[0054] Return to reference Figure 1A For example, process 100 may include applying input data associated with an image (such as that represented by image data 110) and cue points (such as that represented by point data 112) to model 102. Processing the input data, at least based on model 102, process 100 may include model 102 generating and / or outputting feature data 114 representing information associated with features represented by the image. As described herein, in some examples, the information may include, but is not limited to, output points associated with features, heatmaps indicating the location of features represented by the image, and / or representations (e.g., lines) indicating the location of features within a map.

[0055] To learn more details, Figure 1BAn example architecture of a model 102 for detecting features according to some embodiments of the present disclosure is shown. As illustrated, model 102 may include one or more image encoders 116 configured to process image data 110 (e.g., top-down images and / or intensity images) to generate one or more image embeddings 118 associated with the image. As described herein, image encoder 116 may include any type of encoder trained to generate the image embeddings 118 associated with the image, such as a visual transducer. Furthermore, the size of the image embeddings 118 may be based on the size of the image.

[0056] Model 102 may also include one or more cue encoders 120 configured to process point data 112 (e.g., cue points) to generate one or more input tokens 122 (e.g., input cue tokens) associated with one or more cue points. As described herein, cue encoder 120 may include any type of encoder trained to generate input cue tokens 122 associated with cue points. Furthermore, input cue tokens 122 may represent positional codes associated with cue points. Therefore, the number of input cue tokens 122 generated by cue encoder 120 may depend on the number of cue points represented by point data 112. For example, cue encoder 120 may generate a corresponding input cue token 122 for one or more (e.g., each) cue points.

[0057] Furthermore, in some examples, at least one of the input labels 122 may also include learnable labels from the training model 102. In this way, model 102 can append (e.g., concatenate, etc.) the input cue labels 122 associated with the cue point to the learnable input labels 122 to generate the actual labels 122 for further processing by model 102. In some examples, the learnable input labels 122 can help improve the performance of model 102 because the learnable input labels 122 may be image-independent (e.g., avoiding image bias) and / or perform global context aggregation from image features.

[0058] While this example describes input marker 122 as including input cue marker 122 and learnable input marker 122, in other examples (as shown by the dashed lines illustrating optional data and / or components), input marker 122 may include learnable input marker 122 without including input cue marker 122. For example, input marker 122 may include a set of learnable markers learned by model 102 during training. In such an example, model 102 may not receive point data 112 and / or may not include a cue encoder 120 configured to generate input cue marker 122. In other words, model 102 may be configured to perform one or more of the processes described herein without receiving any input from the user.

[0059] In such an example, the number of learnable tags 122 may depend on one or more factors, such as the number of features that can be represented by the image. For example, the number of learnable tags 122 may be equal to or greater than the maximum number of possible features represented by the image. This is because, in some examples, a single learnable tag 122 may be used to query a single feature represented by the image. Therefore, each feature may require a corresponding learnable tag 122.

[0060] Model 102 may also include one or more cross-attention components 124 configured to process image embeddings 118 and input labels 122. As described herein, cross-attention components 124 may include one or more cross-attention layers (e.g., one or more bidirectional cross-attention layers), one or more label-image converters, and / or any other type of processing component configured to perform one or more of the processes described herein. For example, at least based on this processing, cross-attention components 124 may be configured to compare text features associated with cue points with image features associated with images to learn the relationship between text features and image features. Thus, in some examples, the dimension associated with image embeddings 118 may be similar to the dimension associated with input labels 122.

[0061] To learn more details, Figure 1C Examples of cross-attention layers that can be used by one or more machine learning models according to some embodiments of the present disclosure are shown. As shown, cross-attention component 124 may include one or more cross-attention layers 126 and / or one or more cross-attention layers 128. For example, cross-attention component 124 may include two cross-attention layers 126 (and / or any other number of cross-attention layers 126), followed by four cross-attention layers 128 (and / or any other number of cross-attention layers 128). As shown, cross-attention layer 126 may perform pairwise self-attention between query tokens and bidirectional cross-attention between a set of query tokens and a set of image tokens. Next, cross-attention layer 128 may independently perform self-attention of query tokens and bidirectional cross-attention between copies of image tokens and query tokens. Thus, cross-attention layer 126 may share all query tokens and the same image features, while cross-attention layer 128 may include updated image tokens for each query token. More specifically, in some examples, each query tag in cross-attention layer 128 can be cross-attentioned independently with image tags, while all query tags in cross-attention layer 126 can be cross-attentioned with image tags.

[0062] In some examples, cross-attention layers 126 and / or 128 may be associated with one or more of the examples described herein. For instance, cross-attention layers 126 and / or 128 may be used when model 102 is able to automatically perform one or more of the processes described herein, for example, without the need for point data 112 (e.g., model 102 uses only learnable labels). In other words, cross-attention layers 126 and / or 128 can help model 102 generate representations of features without user input.

[0063] Return to reference Figure 1B For example, model 102 may include a cross-attention component 124 that generates and / or outputs one or more output tags 130 and / or one or more image tags 132. As described herein, the output tags 130 may be associated with location information corresponding to cue points represented by point data 112. Furthermore, in some examples, the cross-attention component 124 may be configured to generate and / or output a single output tag 130 associated with each feature. For example, even if the point data 112 represents multiple cue points of a feature, causing the cue encoder 120 to generate multiple input tags 122, the cross-attention component 124 may still be configured to generate a single output tag 130 representing location information associated with the cue points.

[0064] Model 102 may also include one or more decoders 134 configured to process output labels 130 and / or image labels 132 to generate and / or output point data 136 representing one or more points associated with features. As described herein, in some examples, point data 136 may represent a given number of points, such as 100 points (and / or any other number of points), to which decoder 134 is configured and / or trained. Furthermore, point data 136 may represent points using coordinates (e.g., x-coordinate and y-coordinate positions associated with an image and / or map), and / or using sequences. Additionally, in some examples, point data 136 may represent additional information associated with the points, such as the classification associated with the points. As described herein, classification may include, but is not limited to, valid points (e.g., points associated with features), invalid points (e.g., points not associated with features that may be discarded), joint points (e.g., points associated with multiple features), and / or any other type of classification.

[0065] To learn more details, Figure 1D An example is shown where model 102 outputs information associated with points of features according to some embodiments of the present disclosure. In the example... Figure 1DIn this process, decoder 134 may first process a first output label 130(1) (and, in some examples, an image label 132) generated by cross-attention component 124 to generate first point data 136(1) representing a first point associated with a feature. As shown, the first point may be associated with a first x-coordinate position 138(1), a first y-coordinate position 140(1), and a first category 142(1). Then, decoder 134 may process a second output label 130(2) corresponding to the first point (and, in some examples, an image label 132 and / or the first output label 130(1)) to generate second point data 136(2) representing a second point associated with a feature. As shown, the second point may be associated with a second x-coordinate position 138(2), a second y-coordinate position 140(2), and a second category 142(2).

[0066] Then, decoder 134 may process the third output marker 130(3) corresponding to the second point (and, in some examples, image marker 132 and / or output marker 130(1)-(2)) to generate third point data 136(3) representing the third point associated with the feature. As shown, the third point may be associated with the third x-coordinate position 138(3), the third y-coordinate position 140(3), and the third classification 142(3). Decoder 134 may then continue to perform these processes until decoder 134 processes the final output marker 130(N) corresponding to the penultimate point (and, in some examples, image marker 132 and / or any other preceding output marker) to generate final point data 136(N) representing the final point associated with the feature. As shown, the final point may be associated with the final x-coordinate position 138(N), the final y-coordinate position 140(N), and the final classification 142(N).

[0067] As described herein, decoder 134 can be configured to generate point data 136 representing a given number of points (e.g., 100 points (and / or any other number of points)). However, in practice, only a subset of the points may be associated with the feature, such as 40 points (and / or any other number of points). Therefore, classifications 142(1)-(N) (also referred to in the singular as “classification 142” or in the plural as “multiple classifications 142”) can indicate which points are associated with the feature and which points are not. For example, points associated with valid point classifications and / or joint point classifications may be associated with the feature, while points associated with invalid point classifications may not be associated with the feature (e.g., these points may be discarded).

[0068] Furthermore, as shown in the figure, decoder 134 can output points in a sequence, including a first point, followed by a second point, then a third point, and so on, until the final point. Therefore, in some examples, the points being output can continue to be associated with valid classifications and / or joint classifications until a point is associated with an invalid classification. In this instance, the remaining points can then be associated with invalid classifications, causing these points to be unassociated with features and / or discarded. As described in more detail herein, the points, in their output sequence, can be used to generate a final representation associated with the features.

[0069] Although Figure 1D The example shows the output of a single x-coordinate position 138(1)-(N) and a single y-coordinate position 140(1)-(N) for each point, but in other examples, the decoder 134 may output multiple x-coordinate positions and / or multiple y-coordinate positions based at least on the size of the input image. For example, for a point, the decoder 134 may first determine a first number of x-coordinate positions based on the size of the x-coordinate direction of the input image and a second number of y-coordinate positions based on the size of the y-coordinate direction of the input image. For example, if the size of the input image is 1024*512, the decoder 134 may determine that there are 1024 classes in the x-coordinate direction (e.g., one class per pixel) and 512 classes in the y-coordinate direction (e.g., one class per pixel). The decoder 134 (and / or another processing component) can then use the classification to determine the final location associated with that point.

[0070] For example, decoder 134 can select the threshold number of results for the highest x-coordinate class and the threshold number of results for the highest y-coordinate class. As described herein, the threshold number can include any number, such as three results for the x-coordinate class and three results for the y-coordinate class. Furthermore, the highest coordinate class can include the class associated with the highest probability. Decoder 134 can then determine the number of candidate locations using at least the highest x-coordinate class and the highest y-coordinate class. For example, if decoder 134 determines three highest x-coordinate classes and three highest y-coordinate classes, decoder 134 can determine nine candidate locations. Using these candidate locations, decoder 134 then determines the final location associated with that point as including candidate locations associated with the highest overall probability. Decoder 134 can then perform a similar process for one or more other points (e.g., each point) associated with that feature.

[0071] For the representation of points Figure 4 Examples of determining points associated with features represented by image 208 according to some embodiments of this disclosure are shown. Figure 4In the example, image 402 may correspond to image 208, for example, including RGB image 402 associated with portion 206(1) of map 202. As shown, for feature 204(1), decoder 134 may determine a sequence of points 404(1)-(O), which includes a first point 404(1), followed by a second point 404(2), followed by a third point 404(3), and so on (although not all are marked for clarity), until the final point 404(O) (e.g., the last valid point and / or joint point). Decoder 134 may then perform a similar process to determine a point 406 associated with the second feature 204(2) (although only one is marked for clarity) and a point 408 associated with the third feature 204(3) (although only one is marked for clarity).

[0072] Return to reference Figure 1B For example, model 102 may include one or more convolutional layers 144 configured to process output labels 130 and / or image labels 132 to generate and / or output line data 146 representing lines associated with features. As described herein, in some examples, line data 146 may represent a heatmap indicating the location of features represented by an image. For example, if the features include road signs, the heatmap may indicate lines representing the location of road signs within the image. However, in other examples, line data 146 may represent any other type of representation associated with features, such as a color image and / or intensity image indicating the location of features associated with the input image. Furthermore, in other examples, other types of layers in model 102 may be used to generate line data 146.

[0073] For example, Figure 5 An example of a heatmap 502 according to some embodiments of the present disclosure is shown, which indicates the positions of features 204(1)-(3) represented by image 208. As shown, convolutional layer 144 can generate heatmap 502, which includes at least: a first representation 504(1) indicating the position of a first feature 204(1) in image 208; a second representation 504(2) indicating the position of a second feature 204(2) in image 208; and a third representation 504(3) indicating the position of a third feature 204(3) in image 208. Figure 5 In the example, since features 204(1)-(3) include road signs located within the environment, representation 504(1)-(3) can include lines. However, in other examples, representations of other types of features can include any other shape that best represents the features depicted in the image.

[0074] Return to reference Figure 1AFor example, as described above with respect to features, feature data 114 may represent point data 136 (which represents points associated with the features represented by the input image) and / or line data 146 representing a heatmap of the features. Therefore, process 100 may include one or more processing components 148 (in some examples, which may be part of model 102, such as one or more additional layers) configured to process feature data 114 to determine a final representation associated with the features. For example, if the features include road signs, processing component 148 may be configured to determine the final line representing the road signs using at least the points associated with the road signs and the heatmap associated with the road signs.

[0075] For example, since decoder 134 can be configured to determine points in a sequence, processing component 148 can be configured to connect points at least in the order associated with the sequence. Furthermore, processing component 148 can be configured to connect specific points, such as points classified as valid points or points classified as joint points. Additionally, in some examples, processing component 148 can use heatmaps to better connect these points. Processing component 148 can then be configured to generate and / or output final feature data 150, which represents a final representation of the features represented by the input image.

[0076] In some examples, and as by Figure 1A As further illustrated in the example, process 100 may include: one or more classification components 152 processing at least a portion of map data 106 and / or image data 110 to generate classification data 154 representing the classification associated with the feature. As described herein, classification component 152 may include and / or use one or more machine learning models, one or more neural networks, one or more algorithms, one or more classifiers, one or more modules, and / or any other type of processing component to determine the classification associated with the feature. Furthermore, the classification associated with the feature may at least indicate the type associated with the feature. For example, if the feature includes road signs, the classification may include solid line road signs, dashed line road signs, double line road signs, center road signs, two-way road signs, overtaking road signs, arrows, stop lines, pedestrian crossings, and / or any other type of road sign that may be present in the environment.

[0077] Although Figure 1A The example shows classification component 152 as separate from model 102, but in other examples, classification component 152 may include at least a portion of model 102. For example, classification component 152 may include one or more layers of model 102, such that model 102 is further trained to generate classification data 154 representing classifications associated with features.

[0078] In some examples, process 100 may include one or more annotation components 156 annotating a map represented by map data 106 using at least a portion of the final feature data 150 and / or at least a portion of the classification data 154. For example, annotation component 156 may annotate the map to include at least labels for feature locations, classifications associated with the features, and / or any other information associated with the features. Thus, by performing process 100, a map can be automatically annotated with little or no input from one or more users.

[0079] For example, Figure 6 Examples of annotating map 202 to include information associated with feature 204 according to some embodiments of the present disclosure are shown. As shown, annotation component 156 may annotate map 202 using final feature data representing the location of feature 204 and / or classification data representing the classification associated with feature 204. For example, annotation component 156 may annotate map 202 to include at least corresponding labels 602(1)-(8) associated with feature 204 (also referred to in the singular as “label 602” or in the plural as “multiple labels 602”). As described herein, label 602 may at least indicate the location of feature 204 and / or the classification associated with feature 204.

[0080] In some examples, model 102 can be trained to perform one or more processes described herein. For example, Figure 7 A data flow diagram illustrating a process 700 for training model 102 to detect features associated with an image and / or map, according to some embodiments of the present disclosure, is shown. As shown, model 102 can be trained using training input data 702. In some examples, training input data 702 may include image data, point data, image embeddings, and / or input tags. For example, training input data 702 may resemble and / or include image data 110, point data 112, image embeddings 118, and / or input tags 122.

[0081] Model 102 can be trained using training input data 702 and corresponding ground truth data 704. As shown, ground truth data 704 can at least represent a point 706 associated with a feature from the training input data, a line representation 708 associated with the feature, and / or a classification 710 associated with the feature. For example, in some examples, point 706 can indicate the coordinate location associated with a point located on a feature. Furthermore, in some examples, line representation 708 can include a mask and / or heatmap indicating the location of the feature. Furthermore, in some examples, classification 710 can indicate whether point 706 includes valid points, invalid points, joint points, and / or any other point classification. As described herein, ground truth data 704 can be synthetically generated (e.g., generated from a computer model or render), realistically generated (e.g., designed and generated from real-world data), machine-automated (e.g., using feature analysis and learning to extract features from data and then generate labels), human-annotated (e.g., an annotator or annotation expert defining the location of the labels), and / or a combination thereof. In some examples, for each instance of the training input data 702, there may be a corresponding ground truth data 704.

[0082] like Figure 7 As further illustrated, one or more training engines 712 may use one or more loss functions to measure the loss (e.g., error) of the output 714 compared to the ground truth data 704. As shown, the output 714 may also include predicted points 716, predicted line representations 718, and / or predicted classifications 720. Any type of loss function may be used, such as cross-entropy loss, mean squared error, mean absolute error, mean bias error, line splitting loss, and / or other loss function types. In some examples, different outputs 714 may have different loss functions. For example, predicted points 716 may include a first loss function, predicted line representations 718 may include a second loss function, and / or predicted classifications 720 may include a third loss function. In such examples, loss functions may be combined to form a total loss, and the total loss may be used to train model 102 (e.g., update the parameters of model 102). In any example, backpropagation computation may be performed to recursively compute the gradients of the loss functions with respect to the training parameters. In some examples, these gradients may be computed using the weights and biases of model 102.

[0083] Figure 8Examples of one or more systems 802, configured to perform at least a portion of the processes described herein, according to some embodiments of the present disclosure, are shown. As shown, system 802 may include at least one or more processors 804, one or more network interfaces 806, and memory 808. In some examples, processor 804 may include and / or be similar to CPU 1118, GPU 1120, CPU 1206, and / or GPU 1208. Furthermore, network interface 806 may include and / or be similar to network interface 1124 and / or communication interface 1210. Additionally, memory 808 may include and / or be similar to memory 1204.

[0084] The memory 808 may store the model 102, map data 106, segmentation component 104, processing component 148, classification component 152, and / or annotation component 156. Furthermore, the processor 804 may be configured to execute the model 102, segmentation component 104, processing component 148, classification component 152, and / or annotation component 156 to perform one or more of the processes described herein.

[0085] As by Figure 8 As further illustrated in the example, system 802 can communicate with user equipment 108. For example, system 802 can at least send image data 110 to user equipment 108 so that user equipment 108 can present an image to a user. After receiving user input, user equipment 108 can then send point data 112 back to system 802. In other words, system 802 can communicate with user equipment 108 to at least determine and perform... Figure 1A The process is associated with 100 initial prompts.

[0086] In some examples, model 102 can be packaged as a microservice (e.g., an inference microservice (e.g., NVIDIA NIM)), which may include a container (e.g., an operating system (OS)-level virtualization package), which may include an application programming interface (API) layer, a server layer, a runtime layer, and / or a model "engine". For example, the inference microservice may include the container itself and the model (e.g., weights and biases). In some cases, such as when model 102 is small enough (e.g., has a sufficiently small number of parameters), the model may be included within the container itself. In some embodiments, model 102 described herein can be deployed as an inference microservice to accelerate the deployment of models on any cloud, data center, or edge computing system while ensuring data security. For example, an inference microservice may include one or more APIs, pre-configured containers for simplified deployment, an optimized inference engine (e.g., execution software built using standardized AI models, such as NVIDIA's Triton Inference Server), and / or one or more APIs for high-performance deep learning inference, which may include inference runtime and model optimizations that provide low latency and high throughput for production applications (e.g., NVIDIA's TensorRT), and / or enterprise management data for telemetry (e.g., including identity, metrics, health checks, and / or monitoring). The model 102 described herein may be included as part of a microservice along with an acceleration infrastructure capable of deployment using a single command and / or orchestration and autoscaling using a container orchestration system on the acceleration infrastructure (e.g., reaching data center scale on a single device). Therefore, an inference microservice may include machine learning models (e.g., optimized for high-performance inference), inference runtime software for executing model 102 and providing output / response to inputs (e.g., user queries, prompts, etc.), and enterprise management software for providing health checks, identity, and other monitoring. In some embodiments, the inference microservice may include software for performing in-situ replacements and / or updates to model 102. When replacing or updating, the software performing the replacement / update can maintain the user configurations of the inference runtime software and the enterprise management software.

[0087] Now for reference Figure 9 and Figure 10Each block of methods 900 and 1000 described herein includes a computational process that can be executed using any combination of hardware, firmware, and / or software. For example, various functions can be performed by a processor executing instructions stored in memory. Methods 900 and 1000 can also be embodied as computer-usable instructions stored on a computer storage medium. Methods 900 and 1000 can be provided by a standalone application, service, or managed service (independently or in combination with another managed service) or plug-in to another product, to name a few. Furthermore, these methods 900 and 1000 are described by way of example regarding... Figure 1A-1C The methods 900 and 1000 are described herein. However, these methods 900 and 1000 may be additionally or alternatively performed by any system or combination of systems, including but not limited to the systems described herein.

[0088] Figure 9 A flowchart of a method 900 for detecting features associated with an image according to some embodiments of the present disclosure is shown. At block B902, the method may include generating one or more embeddings associated with an image representing the features. For example, an image encoder 116 may process image data 110 representing an image to generate an image embedding 118 associated with the image. As described herein, in some examples, the image may include a portion of a map that is being annotated to include information about the features. However, in other examples, the image may include any other image, such as an image generated using a machine moving in an environment.

[0089] At box B904, method 900 may include generating one or more input labels associated with one or more points corresponding to features. For example, input label 122 associated with features represented by the image may be generated, wherein input label 122 is associated with a first point. As described herein, in some examples, cue encoder 120 may use point data 112 representing the location of the first point within the image to generate at least a portion of input label 122. Furthermore, in some examples, at least a portion of input label 122 may include one or more learnable labels learned during training of model 102. In examples where label 122 includes both input cue label 122 and learnable label 122, input cue label 122 may be appended to learnable label 122.

[0090] At box B906, method 900 may include: using one or more machine learning models to generate output data representing one or more second points corresponding to features, based at least on one or more embeddings and one or more input labels. For example, cross-attention component 124 may first process image embedding 118 and input labels 122 to generate at least output label 130 and image label 132. Decoder 134 may then process output label 130 and / or image label 132 to generate point data 136 representing second points associated with features. In some examples, convolutional layer 144 may further process output label 130 and / or image label 132 to generate line data 146 representing a heatmap of features.

[0091] At box B908, method 900 may include determining a representation of a feature based on at least one or more second points. For example, processing component 148 may process point data 136 and / or line data 146 (e.g., feature data 114) to generate final feature data 150 representing a feature. For example, processing component 148 may generate the representation at least by connecting the second points according to a sequence-based order (where the second points are determined by decoder 134). As described herein, in some examples, if the feature includes road signs, the representation may include a line representation of the road signs.

[0092] At box B910, method 900 may include performing one or more operations based at least on the representation. For example, in some examples, annotation component 156 may annotate the map using at least the final feature data 150 to indicate information associated with the features (e.g., the location of the features, the classification of the features, etc.). However, in other examples, one or more additional and / or alternative processes may be performed, such as enabling the machine to operate at least based on the representation and / or location of the features.

[0093] Figure 10 A flowchart of another method 1000 for detecting features associated with an image according to some embodiments of the present disclosure is shown. At block B1002, method 1000 may include: determining one or more first points associated with features represented by an image corresponding to a map. For example, the first points associated with features represented by the image may be determined. As described herein, in some examples, the first points may be determined based at least on one or more user inputs indicating the first points. Additionally or alternatively, in some examples, the first points may be automatically determined using one or more learnable markers 122.

[0094] At box B1004, method 1000 may include: using one or more machine learning models to determine one or more second points associated with features based at least on input data associated with one or more first points and images. For example, model 102 may process input data representing first points and images. In some examples, this process may include generating input labels 122 using the first points, generating image embeddings 118 using the images, generating output labels 130 and / or image labels 132 using a cross-attention component 124, and then generating point data 136 representing the second points using a decoder 134.

[0095] At box B1006, method 1000 may include generating a representation of a feature based on at least one or more second points. For example, processing component 148 may process point data 136 (e.g., feature data 114) to generate final feature data 150 representing the feature. For example, processing component 148 may generate the representation at least by connecting the second points according to a sequence-based order (which determines the second points of the sequence). As described herein, in some examples, if the feature includes road signs, the representation may include a line representation of the road signs.

[0096] At box B1008, method 1000 may include updating the map, at least based on the representation, to include information associated with the feature. For example, annotation component 156 may update the map to include information associated with the feature. As described herein, in some examples, this information may include at least the location of the feature, the classification of the feature, and / or any other information associated with the feature.

[0097] Example autonomous vehicles

[0098] Figure 11AThis is an illustration of an example autonomous vehicle 1100 according to some embodiments of the present disclosure. The autonomous vehicle 1100 (or, alternatively, referred to herein as “vehicle 1100”) may include, but is not limited to, passenger vehicles such as cars, trucks, buses, first-response vehicles, shuttles, electric or motorized bicycles, motorcycles, fire trucks, police vehicles, ambulances, boats, engineering vehicles, submarines, robotic vehicles, drones, aircraft, vehicles coupled to trailers (e.g., semi-trailer trucks for hauling goods) and / or other types of vehicles (e.g., driverless and / or vehicles accommodating one or more passengers). Autonomous vehicles are typically described according to the levels of automation defined by the National Highway Traffic Safety Administration (NHTSA), a division of the U.S. Department of Transportation, and the Society of Automotive Engineers (SAE) in its "Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles" (Standard No. J3016-201806, published June 15, 2018; Standard No. J3016-201609, published September 30, 2016; and previous and future versions of this standard). Vehicle 1100 may be able to implement one or more functions that meet Level 3-5 of the autonomous driving level. Vehicle 1100 may be able to implement one or more functions that meet Level 1-5 of the autonomous driving level. For example, depending on the embodiment, vehicle 1100 may be able to implement driver assistance (Level 1), partial automation (Level 2), conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5). As used herein, the term “autonomy” can include any and / or all types of autonomy of the vehicle 1100 or other machine, such as full autonomy, high autonomy, conditional autonomy, partial autonomy, assisted autonomy, semi-autonomy, primary autonomy or other designations.

[0099] Vehicle 1100 may include components such as chassis, body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other vehicle components. Vehicle 1100 may include a propulsion system 1150, such as an internal combustion engine, a hybrid power plant, an all-electric motor, and / or another type of propulsion system. Propulsion system 1150 may be connected to the drivetrain of vehicle 1100, which may include a transmission, to enable propulsion of vehicle 1100. Propulsion system 1150 may be controlled in response to receiving a signal from throttle / accelerator 1152.

[0100] A steering system 1154, which may include a steering wheel, can be used to steer the vehicle 1100 (e.g., along a desired path or route) when the propulsion system 1150 is operating (e.g., when the vehicle is in motion). The steering system 1154 may receive signals from the steering actuator 1156. For fully automatic (level 5) functionality, the steering wheel may be optional.

[0101] The brake sensor system 1146 can be used to operate the vehicle brakes in response to receiving signals from the brake actuator 1148 and / or the brake sensor.

[0102] It may include one or more System-on-Chip (SoC) 1104 ( Figure 11C One or more controllers 1136, including and / or one or more GPUs, may provide signals (e.g., signals representing commands) to one or more components and / or systems of vehicle 1100. For example, one or more controllers may send signals to operate vehicle brakes via one or more brake actuators 1148, to operate steering system 1154 via one or more steering actuators 1156, and to operate propulsion system 1150 via one or more throttles / accelerators 1152. One or more controllers 1136 may include one or more onboard (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and output operating commands (e.g., signals representing commands) to enable autonomous driving and / or assist a human driver in driving vehicle 1100. One or more controllers 1136 may include a first controller 1136 for autonomous driving functions, a second controller 1136 for functional safety functions, a third controller 1136 for artificial intelligence functions (e.g., computer vision), a fourth controller 1136 for infotainment functions, a fifth controller 1136 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 1136 can handle two or more of the functions described above, and two or more controllers 1136 can handle a single function, and / or any combination thereof.

[0103] One or more controllers 1136 may provide signals for controlling one or more components and / or systems of vehicle 1100 in response to sensor data (e.g., sensor inputs) received from one or more sensors. Sensor data may be received from, for example, but not limited to, Global Navigation Satellite System (“GNSS”) sensors 1158 (e.g., Global Positioning System sensors), RADAR sensors 1160, ultrasonic sensors 1162, LIDAR sensors 1164, inertial measurement unit (IMU) sensors 1166 (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), microphones 1196, stereo cameras 1168, wide-angle cameras 1170 (e.g., fisheye cameras), infrared cameras 1172, surround cameras 1174 (e.g., 360-degree cameras), long-range and / or medium-range cameras 1198, speed sensors 1144 (e.g., for measuring the rate of vehicle 1100), vibration sensors 1142, steering sensors 1140, braking sensors (e.g., as part of braking sensor system 1146), and / or other sensor types.

[0104] One or more of the controllers 1136 may receive inputs (e.g., represented by input data) from the instrument cluster 1132 of the vehicle 1100 and provide outputs (e.g., represented by output data, display data, etc.) via a human-machine interface (HMI) display 1134, an auditory signaling device, a speaker, and / or via other components of the vehicle 1100. These outputs may include information such as vehicle speed, rate, time, map data (e.g., [missing information]). Figure 11C Information such as a high-definition (“HD”) map 1122, location data (e.g., the location of vehicle 1100 on the map), direction, the location of other vehicles (e.g., occupying a grid), and information about objects and their states perceived by controller 1136, etc. For example, HMI display 1134 may display information about the existence of one or more objects (e.g., street signs, warning signs, traffic light changes, etc.) and / or information about driving maneuvers that the vehicle has made, is making, or will make (e.g., changing lanes now, leaving 34B in two miles, etc.).

[0105] Vehicle 1100 also includes a network interface 1124, which can communicate via one or more networks using one or more wireless antennas 1126 and / or a modem. For example, network interface 1124 may be able to communicate via Long Term Evolution (“LTE”), Wideband Code Division Multiple Access (“WCDMA”), Universal Mobile Telecommunications System (“UMTS”), Global System for Mobile Communications (“GSM”), IMT-CDMA Multicarrier (“CDMA2000”), etc. One or more wireless antennas 1126 may also enable communication between objects in the environment (e.g., vehicles, mobile devices, etc.) using one or more local area networks such as Bluetooth, Bluetooth Low Energy (“LE”), Z-Wave, ZigBee, etc., and / or one or more low-power wide area networks (LPWANs such as LoRaWAN, SigFox, etc.).

[0106] Figure 11B For use in accordance with some embodiments of this disclosure Figure 11A This is an example of the camera position and field of view of an example autonomous vehicle 1100. The camera and its respective field of view are an example embodiment and are not intended to be limiting. For example, additional and / or replaceable cameras may be included, and / or these cameras may be located at different positions on the vehicle 1100.

[0107] The camera type used for the camera may include, but is not limited to, a digital camera suitable for use with components and / or systems of vehicle 1100. The camera may operate at Automotive Safety Integrity Level (ASIL) B and / or another ASIL. The camera type may have any image capture rate, such as 60 frames per second (fps), 120 fps, 240 fps, etc., depending on the embodiment. The camera may be able to use a rolling shutter, a global shutter, another type of shutter, or a combination thereof. In some examples, the color filter array may include a red-white-white-white (RCCC) color filter array, a red-white-white-blue (RCCB) color filter array, a red-blue-green-white (RBGC) color filter array, a Foveon X3 color filter array, a Bayer sensor (RGGB) color filter array, a monochrome sensor color filter array, and / or another type of color filter array. In some embodiments, a sharp-pixel camera, such as a camera with RCCC, RCCB, and / or RBGC color filter arrays, may be used in efforts to improve light sensitivity.

[0108] In some examples, one or more of the cameras can be used to perform advanced driver assistance system (ADAS) functions (e.g., as part of a redundant or fail-safe design). For example, a multi-function monocular camera can be installed to provide functions including lane departure warning, traffic sign assistance, and intelligent headlight control. One or more of the cameras (e.g., all cameras) can simultaneously record and provide image data (e.g., video).

[0109] One or more of the cameras can be mounted in mounting components such as custom-designed (3D-printed) components to cut off stray light and reflections from inside the vehicle (e.g., reflections from the dashboard reflected in the windshield mirror) that may interfere with the camera's image data capture capabilities. Regarding the wing mirror mounting components, the wing mirror components can be custom-3D printed so that the camera mounting plate matches the shape of the wing mirror. In some examples, one or more cameras can be integrated into the wing mirror. For side-view cameras, one or more cameras can also be integrated into the four pillars at each corner of the cab.

[0110] A camera with a field of view that includes the environment in front of the vehicle 1100 (e.g., a front-facing camera) can be used for surround view to help identify forward paths and obstacles, and, with the assistance of one or more controllers 1136 and / or control SoCs, to provide information crucial for generating an occupancy grid and / or determining a preferred vehicle path. The front-facing camera can be used to perform many of the same ADAS functions as LiDAR, including emergency braking, pedestrian detection, and collision avoidance. The front-facing camera can also be used in ADAS functions and systems, including Lane Departure Warning (“LDW”), Autonomous Cruise Control (“ACC”), and / or other functions such as traffic sign recognition.

[0111] A variety of cameras can be used in front-facing configurations, including, for example, monocular camera platforms including complementary metal-oxide-semiconductor (“CMOS”) color imagers. Another example could be a wide-angle camera 1170, which can be used to perceive objects entering the field of view from the periphery (such as pedestrians, traffic at intersections, or bicycles). Although Figure 11B The diagram shows only one wide-angle camera, but any number (including zero) of wide-angle cameras 1170 can be present on vehicle 1100. Furthermore, any number of one or more remote cameras 1198 (e.g., a pair of long-view stereo cameras) can be used for depth-based object detection, especially for objects for which a neural network has not yet been trained. One or more remote cameras 1198 can also be used for object detection and classification, as well as basic object tracking.

[0112] Any number of stereo cameras 1168 may also be included in a front-mounted configuration. In at least one embodiment, one or more stereo cameras 1168 may include an integrated control unit comprising a scalable processing unit that can provide a multi-core microprocessor and programmable logic (“FPGA”) with an integrated controller area network (“CAN”) or Ethernet interface on a single chip. Such a unit can be used to generate a 3D map of the vehicle environment, including distance estimates for all points in the image. Alternative stereo cameras 1168 may include a compact stereo vision sensor that may include two camera lenses (one on each side) and an image processing chip capable of measuring the distance from the vehicle to a target object and using the generated information (e.g., metadata) to activate autonomous emergency braking and lane departure warning functions. Other types of stereo cameras 1168 may be used in addition to those described herein or alternatively.

[0113] A camera (e.g., a side-view camera) having a field of view that includes the side of the vehicle 1100 can be used for surround view, providing information for creating and updating occupancy grids and generating side-impact collision warnings. For example, a surround camera 1174 (e.g., ...) Figure 11B The four surround cameras 1174 shown can be mounted on vehicle 1100. The surround cameras 1174 can include wide-angle cameras 1170, fisheye cameras, 360-degree cameras, and / or the like. Four examples are provided; the four fisheye cameras can be positioned at the front, rear, and sides of the vehicle. In an alternative arrangement, the vehicle can use three surround cameras 1174 (e.g., left, right, and rear) and can utilize one or more other cameras (e.g., forward-facing cameras) as a fourth surround-view camera.

[0114] A camera (e.g., a rear-view camera) having a field of view that includes the environment behind the vehicle 1100 can be used for parking assistance, surround view, rear collision warning, and creating and updating occupancy grids. A wide variety of cameras can be used, including but not limited to those also suitable as front-facing cameras as described herein (e.g., long-range and / or mid-range camera 1198, stereo camera 1168, infrared camera 1172, etc.).

[0115] Figure 11C For use in accordance with some embodiments of this disclosure Figure 11AThe example autonomous vehicle 1100 is illustrated in the block diagram of an example system architecture. It should be understood that this arrangement, and other arrangements described herein, are merely illustrative. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of those shown, and some elements may be omitted entirely. Furthermore, many of the elements described herein are functional entities, which may be implemented as discrete or distributed components or in combination with other components, and in any suitable combination and location. The various functions described herein as being performed by these entities can be implemented via hardware, firmware, and / or software. For example, the various functions can be implemented by a processor executing instructions stored in memory.

[0116] Figure 11C Each component, feature, and system in vehicle 1100 is illustrated as being connected via bus 1102. Bus 1102 may include a Controller Area Network (CAN) data interface (or, alternatively, referred to herein as the "CAN bus"). CAN may be a network within vehicle 1100 used to assist in the control of various features and functions of vehicle 1100, such as the actuation of brakes, acceleration, braking, steering, windshield wipers, etc. The CAN bus may be configured to have dozens or even hundreds of nodes, each with its own unique identifier (e.g., CAN ID). The CAN bus can be read to find steering wheel angle, ground speed, engine speed per minute (RPM), button positions, and / or other vehicle status indicators. The CAN bus may be ASIL B compliant.

[0117] Although bus 1102 is described herein as a CAN bus, this is not intended to be limiting. For example, FlexRay and / or Ethernet may be used in addition to or alternatively to a CAN bus. Furthermore, although bus 1102 is represented by a single line, this is not intended to be limiting. For example, any number of buses 1102 may exist, which may include one or more CAN buses, one or more FlexRay buses, one or more Ethernet buses, and / or one or more other types of buses using different protocols. In some examples, two or more buses 1102 may be used to perform different functions and / or may be used for redundancy. For example, a first bus 1102 may be used for a collision avoidance function, and a second bus 1102 may be used for drive control. In any example, each bus 1102 may communicate with any component of vehicle 1100, and two or more buses 1102 may communicate with the same component. In some examples, each SoC 1104, each controller 1136, and / or each computer within the vehicle may have access to the same input data (e.g., input from sensors in the vehicle 1100) and may be connected to a common bus such as the CAN bus.

[0118] Vehicle 1100 may include one or more controllers 1136, such as those described herein. Figure 11A The controllers described herein. Controller 1136 can be used for a wide variety of functions. Controller 1136 can be coupled to any other different components and systems of vehicle 1100 and can be used for the control of vehicle 1100, artificial intelligence of vehicle 1100, infotainment and / or the like for vehicle 1100.

[0119] Vehicle 1100 may include one or more System-on-Chip (SoC) 1104. SoC 1104 may include CPU 1106, GPU 1108, processor 1110, cache 1112, accelerator 1114, data storage 1116, and / or other components and features not shown. SoC 1104 can be used to control vehicle 1100 across a wide variety of platforms and systems. For example, one or more SoCs 1104 may be combined with an HD map 1122 in a system (e.g., the system of vehicle 1100), the HD map being accessible from one or more servers (e.g., via a network interface 1124). Figure 11D One or more servers (1178) receive map refresh and / or updates.

[0120] CPU 1106 may include CPU clusters or CPU complexes (or, alternatively, referred to herein as "CCPLEX"). CPU 1106 may include multiple cores and / or L2 cache. For example, in some embodiments, CPU 1106 may include eight cores in a coherent multiprocessor configuration. In some embodiments, CPU 1106 may include four dual-core clusters, each cluster having a dedicated L2 cache (e.g., 2MB L2 cache). CPU 1106 (e.g., CCPLEX) may be configured to support simultaneous cluster operation, such that any combination of clusters of CPU 1106 can be active at any given time.

[0121] CPU 1106 can implement power management capabilities including one or more of the following features: automatic clock gating of hardware blocks when idle to save dynamic power; clock gating of each core when the core is not actively executing instructions due to the execution of WFI / WFE instructions; independent power gating of each core; independent clock gating of each core cluster when all cores are clock-gated or power-gated; and / or independent power gating of each core cluster when all cores are power-gated. CPU 1106 can further implement enhanced algorithms for managing power states, wherein allowed power states and desired wake-up times are specified, and the hardware / microcode determines the optimal power state to enter for the core, cluster, and CCPLEX. The processing core can support simplified power state entry sequences in software, with this work offloaded to the microcode.

[0122] GPU 1108 may include an integrated GPU (or, alternatively, referred to herein as an "iGPU"). GPU 1108 may be programmable and efficient for parallel workloads. In some examples, GPU 1108 may use an enhanced tensor instruction set. GPU 1108 may include one or more streaming microprocessors, wherein each streaming microprocessor may include an L1 cache (e.g., an L1 cache with at least 96KB of storage capacity), and two or more of these streaming microprocessors may share an L2 cache (e.g., an L2 cache with 512KB of storage capacity). In some embodiments, GPU 1108 may include at least eight streaming microprocessors. GPU 1108 may use a computation application programming interface (API). Furthermore, GPU 1108 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0123] In automotive and embedded applications, the GPU 1108 can be power-optimized for optimal performance. For example, the GPU 1108 can be fabricated on FinFETs. However, this is not intended to be limiting, and the GPU 1108 can be fabricated using other semiconductor manufacturing processes. Each streaming microprocessor can combine several mixed-precision processing cores divided into multiple blocks. For example, and without limitation, 64 PF32 cores and 32 PF64 cores can be divided into four processing blocks. In such an example, each processing block can be allocated 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed-precision NVIDIA Tensor cores for deep learning matrix arithmetic, an L0 instruction cache, a warp scheduler, a dispatch unit, and / or a 64KB register file. Furthermore, the streaming microprocessor can include independent parallel integer and floating-point data paths to leverage the mixture of computation and addressing computations to provide efficient execution of workloads. Streaming microprocessors may include independent thread scheduling capabilities to allow for finer-grained synchronization and cooperation between parallel threads. Streaming microprocessors may include combined L1 data caches and shared memory units to improve performance while simplifying programming.

[0124] The GPU 1108 may include, in some examples, a High Bandwidth Memory (HBM) and / or a 16GB HBM2 memory subsystem providing a peak memory bandwidth of approximately 900GB / s. In some examples, in addition to HBM memory or alternatively, Synchronous Graphics Random Access Memory (SGRAM), such as Generation 5 Graphics Double Data Rate Synchronous Random Access Memory (GDDR5), may be used.

[0125] The GPU 1108 may include unified memory technology, which includes access counters to allow memory pages to be migrated more precisely to the processors that access them most frequently, thereby improving the efficiency of shared memory ranges between processors. In some examples, Address Translation Service (ATS) support can be used to allow the GPU 1108 to directly access the CPU 1106 page tables. In such examples, when the GPU 1108 Memory Management Unit (MMU) experiences a miss, the address translation request can be transferred to the CPU 1106. In response, the CPU 1106 can look up the virtual-physical mapping for the address in its page tables and transfer the translation back to the GPU 1108. Thus, unified memory technology can allow a single unified virtual address space for the memory of both the CPU 1106 and the GPU 1108, simplifying GPU 1108 programming and porting applications to the GPU 1108.

[0126] In addition, the GPU 1108 may include access counters that track how frequently the GPU 1108 accesses the memory of other processors. Access counters can help ensure that memory pages are moved to the physical memory of the processor that accesses those pages most frequently.

[0127] SoC 1104 may include any number of caches 1112, including those described herein. For example, cache 1112 may include an L3 cache available to both CPU 1106 and GPU 1108 (e.g., it is connected to both CPU 1106 and GPU 1108). Cache 1112 may include a write-back cache, which can track the state of rows, for example, using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). Depending on the embodiment, the L3 cache may include 4 MB or more, but a smaller cache size may also be used.

[0128] SoC 1104 may include an arithmetic logic unit (ALU) that can be utilized in processing of any of the various tasks or operations performed on vehicle 1100, such as processing a DNN. Furthermore, SoC 1104 may include a floating-point unit (FPU) (or other mathematical coprocessor or digital coprocessor type) for performing mathematical operations within the system. For example, SoC 104 may include one or more FPUs integrated as execution units within CPU 1106 and / or GPU 1108.

[0129] SoC 1104 may include one or more accelerators 1114 (e.g., hardware accelerators, software accelerators, or combinations thereof). For example, SoC 1104 may include a hardware accelerator cluster, which may include optimized hardware accelerators and / or large on-chip memory. This large on-chip memory (e.g., 4MB SRAM) can enable the hardware accelerator cluster to accelerate neural networks and other computations. The hardware accelerator cluster can be used to complement GPU 1108 and offload some tasks from GPU 1108 (e.g., freeing up more cycles of GPU 1108 to perform other tasks). As an example, accelerator 1114 can be used for targeted workloads (e.g., perceptrons, convolutional neural networks (CNNs), etc.) that are sufficiently stable to allow for easy control of acceleration. When used herein, the term "CNN" can include all types of CNNs, including region-based or region convolutional neural networks (RCNNs) and fast RCNNs (e.g., for object detection).

[0130] Accelerator 1114 (e.g., a hardware accelerator cluster) may include a Deep Learning Accelerator (DLA). The DLA may include one or more Tensor Processing Units (TPUs) that can be configured to provide an additional 10 trillion operations per second for deep learning applications and inference. The TPU may be an accelerator configured to perform image processing functions (e.g., for CNNs, RCNNs, etc.) and optimized for performing image processing functions. The DLA may be further optimized for a specific set of neural network types and floating-point operations and inference. The DLA is designed to provide higher performance per millimeter than a general-purpose GPU and significantly outperform CPUs. The TPU can perform several functions, including single-instance convolution functions, support for INT8, INT16, and FP16 data types for both features and weights, and post-processor functions.

[0131] DLA can execute neural networks, especially CNNs, quickly and efficiently on processed or unprocessed data for any function across a wide variety of applications, such as, but not limited to: CNNs for object recognition and detection using data from camera sensors; CNNs for distance estimation using data from camera sensors; CNNs for emergency vehicle detection and recognition using data from microphones; CNNs for face recognition and vehicle owner recognition using data from camera sensors; and / or CNNs for safety and / or safety-related events.

[0132] The DLA can perform any function of the GPU 1108, and by using inference accelerators, for example, designers can make the DLA or GPU 1108 target any function. For example, designers can focus the CNN processing and floating-point operations on the DLA and leave other functions to the GPU 1108 and / or other accelerators 1114.

[0133] Accelerator 1114 (e.g., a cluster of hardware accelerators) may include a programmable vision accelerator (PVA), which may alternatively be referred to herein as a computer vision accelerator. The PVA may be designed and configured to accelerate computer vision algorithms for advanced driver assistance systems (ADAS), autonomous driving, and / or augmented reality (AR) and / or virtual reality (VR) applications. The PVA can provide a balance between performance and flexibility. For example, each PVA may include, for example, but not limited to, any number of reduced instruction set computer (RISC) cores, direct memory access (DMA), and / or any number of vector processors.

[0134] RISC cores can interact with image sensors (such as the image sensor of any camera described herein), image signal processors, and / or the like. Each of these RISC cores may include any amount of memory. Depending on the embodiment, the RISC core may use any of several protocols. In some examples, the RISC core may execute a real-time operating system (RTOS). RISC cores may be implemented using one or more integrated circuit devices, application-specific integrated circuits (ASICs), and / or memory devices. For example, a RISC core may include an instruction cache and / or tightly coupled RAM.

[0135] DMA enables PVA components to access system memory independently of the CPU 1106. DMA can support any number of features to provide optimizations to the PVA, including but not limited to support for multidimensional addressing and / or circular addressing. In some examples, DMA can support addressing in up to six or more dimensions, which can include block width, block height, block depth, horizontal block step, vertical block step, and / or depth step.

[0136] A vector processor can be a programmable processor designed to efficiently and flexibly execute programming for computer vision algorithms and provide signal processing capabilities. In some examples, a PVA may include a PVA core and two vector processing subsystem partitions. The PVA core may include a processor subsystem, one or more DMA engines (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem may operate as the main processing engine of the PVA and may include a vector processing unit (VPU), an instruction cache, and / or vector memory (e.g., VMEM). The VPU core may include a digital signal processor, such as, for example, a Single Instruction Multiple Data (SIMD) or Very Long Instruction Word (VLIW) digital signal processor. The combination of SIMD and VLIW can enhance throughput and speed.

[0137] Each of the vector processors may include an instruction cache and may be coupled to dedicated memory. As a result, in some examples, each of the vector processors may be configured to execute independently of other vector processors. In other examples, the vector processors included in a particular PVA may be configured to employ data parallelism. For example, in some embodiments, multiple vector processors included in a single PVA may execute the same computer vision algorithm, but on different regions of an image. In other examples, vector processors included in a particular PVA may execute different computer vision algorithms simultaneously on the same image, or even different algorithms on a sequence of images or portions of an image. Among other things, any number of PVAs may be included in a cluster of hardware accelerators, and any number of vector processors may be included in each of these PVAs. Furthermore, the PVA may include additional error correction code (ECC) memory to enhance overall system security.

[0138] Accelerator 1114 (e.g., a hardware accelerator cluster) may include an on-chip computer vision network and SRAM to provide high-bandwidth, low-latency SRAM for accelerator 1114. In some examples, on-chip memory may include at least 4MB of SRAM consisting of, for example, but not limited to, eight field-configurable memory blocks, accessible by both the PVA and DLA. Each pair of memory blocks may include an Advanced Peripheral Bus (APB) interface, configuration circuitry, a controller, and a multiplexer. Any type of memory may be used. The PVA and DLA may access memory via a backbone that provides high-speed memory access to the PVA and DLA. The backbone may include (e.g., using an APB) an on-chip computer vision network that interconnects the PVA and DLA to memory.

[0139] On-chip computer vision networks can include interfaces that ensure both the PVA and DLA provide ready and valid signals before transmitting any control signals / addresses / data. Such interfaces can provide separate phases and channels for transmitting control signals / addresses / data, as well as burst communication for continuous data transmission. This type of interface can conform to ISO 26262 or IEC 61508 standards, but other standards and protocols can also be used.

[0140] In some examples, SoC 1104 may include, for example, a real-time ray tracing hardware accelerator as described in U.S. Patent Application No. 16 / 101,232, filed August 10, 2018. This real-time ray tracing hardware accelerator can be used to quickly and efficiently determine the location and extent of objects (e.g., within a world model) to generate real-time visualization simulations for RADAR signal interpretation, sound propagation synthesis and / or analysis, SONAR system simulation, general wave propagation simulation, comparison with LiDAR data for localization and / or other functional purposes, and / or other uses. In some embodiments, one or more Tree Traversal Units (TTUs) may be used to perform one or more ray tracing-related operations.

[0141] Accelerators 1114 (e.g., hardware accelerator clusters) have broad applications in autonomous driving. PVAs can be programmable vision accelerators used in critical processing stages of ADAS and autonomous vehicles. The capabilities of PVAs are a good match for algorithmic domains requiring predictable processing, low power, and low latency. In other words, PVAs perform well in semi-dense or dense rule computation, even on small datasets requiring predictable runtimes with low latency and low power. Therefore, in the context of platforms for autonomous vehicles, PVAs are designed to run classical computer vision algorithms because they are effective in object detection and integer mathematical operations.

[0142] For example, according to one embodiment of this technology, PVA is used to perform computer stereo vision. In some examples, semi-global matching-based algorithms may be used, but this is not intended to be limiting. Many applications for Level 3-5 autonomous driving require instantaneous motion estimation / stereo matching (e.g., from moving structures, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions on input from two monocular cameras.

[0143] In some examples, PVA can be used to perform intensive optical flow, providing processed RADAR data from the raw RADAR data (e.g., using 4D Fast Fourier Transform). In other examples, PVA is used for time-of-flight depth processing, which, for example, involves processing raw time-of-flight data to provide processed time-of-flight data.

[0144] DLA can be used to run any type of network to enhance control and driving safety, including, for example, neural networks that output a confidence metric for each object detection. Such a confidence value can be interpreted as a probability or as providing a relative “weight” for each detection compared to other detections. This confidence value allows the system to make further decisions about which detections should be considered true positives rather than false positives. For example, the system can set a threshold for the confidence and only consider detections exceeding the threshold as true positives. In an Automatic Emergency Braking (AEB) system, false positives can cause the vehicle to automatically perform emergency braking, which is clearly undesirable. Therefore, only the most confident detections should be considered as triggers for AEB. DLA can run neural networks to regress the confidence values. The neural network can take at least some subset of parameters as its input, such as bounding box dimensions, ground plane estimates obtained (e.g. from another subsystem), outputs from inertial measurement unit (IMU) sensors 1166 related to the orientation and distance of vehicle 1100, 3D position estimates of objects obtained from the neural network and / or other sensors (e.g., LiDAR sensor 1164 or RADAR sensor 1160), etc.

[0145] SoC 1104 may include one or more data storage units 1116 (e.g., memory). Data storage units 1116 may be on-chip memory of SoC 1104, which may store neural networks to be executed on the GPU and / or DLA. In some examples, for redundancy and security, data storage units 1116 may be large enough to store multiple instances of the neural network. Data storage units 1112 may include L2 or L3 cache 1112. References to data storage units 1116 may include references to memory associated with PVA, DLA, and / or other accelerators 1114 as described herein.

[0146] SoC 1104 may include one or more processors 1110 (e.g., embedded processors). Processor 1110 may include a startup and power management processor, which may be a dedicated processor and subsystem for handling startup power and management functions, as well as safety implementation. The startup and power management processor may be part of the SoC 1104 startup sequence and may provide runtime power management services. The startup power and management processor may provide clock and voltage programming, auxiliary system low-power state transitions, SoC 1104 thermal and temperature sensor management, and / or SoC 1104 power state management. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to the temperature, and SoC 1104 may use the ring oscillator to detect the temperature of CPU 1106, GPU 1108, and / or accelerator 1114. If it is determined that the temperature exceeds a threshold, the startup and power management processor may enter a temperature fault routine and place SoC 1104 into a lower power state and / or place vehicle 1100 into a driver-safe parking mode (e.g., safely stop vehicle 1100).

[0147] The processor 1110 may also include a set of embedded processors that can be used as an audio processing engine. The audio processing engine can be an audio subsystem that allows for full hardware support for multi-channel audio via multiple interfaces and a wide and flexible range of audio I / O interfaces. In some examples, the audio processing engine is a dedicated processor core with a digital signal processor and dedicated RAM.

[0148] The processor 1110 may also include an always-on-processor engine that can provide the necessary hardware features to support low-power sensor management and wake-up use cases. This always-on-processor engine may include a processor core, tightly coupled RAM, support for peripherals (such as timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0149] The processor 1110 may also include a security cluster engine, which includes a dedicated processor subsystem for handling security management for automotive applications. The security cluster engine may include two or more processor cores, tightly coupled RAM, support for peripheral devices (e.g., timers, interrupt controllers, etc.), and / or routing logic. In secure mode, the two or more cores may operate in lockstep mode and function as a single core with comparison logic that detects any differences between their operations.

[0150] The processor 1110 may also include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0151] The processor 1110 may also include a high dynamic range signal processor, which may include an image signal processor, which is a hardware engine that is part of the camera processing pipeline.

[0152] Processor 1110 may include a video image compositer, which may be (e.g., implemented on a microprocessor) a processing block, implementing video post-processing functions required by the video playback application to generate the final image for the player window. The video image compositer may perform lens distortion correction on the wide-angle camera 1170, the surround camera 1174, and / or the in-cabin monitoring camera sensor. The in-cabin monitoring camera sensor is preferably monitored by a neural network running on another instance of an advanced SoC, configured to recognize in-cabin events and respond accordingly. The in-cabin system may perform lip reading to activate mobile phone services and make calls, dictate emails, change vehicle destinations, activate or change the vehicle's infotainment system and settings, or provide voice-activated web browsing. Some functions are only available to the driver when the vehicle is operating in autonomous mode and are disabled in other situations.

[0153] Video image compositers can include enhanced temporal denoising for both spatial and temporal noise reduction. For example, in the case of motion in the video, denoising appropriately weights spatial information, reducing the weight of information provided by neighboring frames. In cases where the image or part of the image does not contain motion, the temporal denoising performed by the video image compositer can use information from previous images to reduce noise in the current image.

[0154] The video image compositer can also be configured to perform stereo correction on input stereo lens frames. When the operating system desktop is in use and the GPU 1108 does not need to continuously render new surfaces, the video image compositer can be further used for user interface components. Even when the GPU 1108 is powered on and active, performing 3D rendering, the video image compositer can be used to offload the GPU 1108 to improve performance and responsiveness.

[0155] The SoC 1104 may also include a Mobile Industry Processor Interface (MIPI) camera serial interface, a high-speed interface, and / or a video input block that can be used for camera and related pixel input functions for receiving video and input from a camera. The SoC 1104 may also include an input / output controller that can be software controlled and can be used to receive I / O signals not assigned to a specific role.

[0156] SoC 1104 may also include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management and / or other devices. SoC 1104 can be used to process data from cameras and sensors (e.g., LIDAR sensor 1164, RADAR sensor 1160, etc., which can be connected via Gigabit Multimedia Serial Link and Ethernet), data from bus 1102 (e.g., vehicle 1100 speed, steering wheel position, etc.), and data from GNSS sensor 1158 (connected via Ethernet or CAN bus). SoC 1104 may also include a dedicated high-performance, high-capacity memory controller, which may include its own DMA engine and can be used to free up CPU 1106 from routine data management tasks.

[0157] The SoC 1104 can be an end-to-end platform with a flexible architecture spanning Automation Levels 3-5, providing a comprehensive functional safety architecture that leverages and efficiently utilizes computer vision and ADAS technologies for diversity and redundancy, along with deep learning tools to deliver a flexible and reliable driving software stack. The SoC 1104 can be faster, more reliable, and even more energy- and space-efficient than conventional systems. For example, when combined with the CPU 1106, GPU 1108, and data storage 1116, the accelerator 1114 can provide a fast and efficient platform for Level 3-5 autonomous vehicles.

[0158] Therefore, this technology offers capabilities and functionalities that cannot be achieved through conventional systems. For example, computer vision algorithms can be executed on CPUs, which can be configured using high-level programming languages ​​such as C to execute a wide variety of processing algorithms across a diverse range of visual data. However, CPUs often cannot meet the performance requirements of many computer vision applications, such as those related to execution time and power consumption. In particular, many CPUs cannot execute complex object detection algorithms in real time, which is a requirement for automotive ADAS applications and practical Level 3-5 autonomous vehicles.

[0159] In contrast to conventional systems, the techniques described in this paper, by providing CPU complexes, GPU complexes, and hardware accelerator clusters, allow multiple neural networks to be executed simultaneously and / or sequentially, and the results combined to achieve Level 3–5 autonomous driving capabilities. For example, a CNN executed on a DLA or dGPU (e.g., GPU 1120) could include text and word recognition, allowing a supercomputer to read and understand traffic signs, including those for which neural networks have not yet been specifically trained. The DLA could also include a neural network capable of recognizing, interpreting, and providing semantic understanding of the signs, and passing that semantic understanding to a path planning module running on the CPU complex.

[0160] As another example, multiple neural networks can operate simultaneously, as required for Level 3, 4, or 5 driving. For instance, a warning sign consisting of "Caution: Flashing lights indicate icy conditions," along with a light, can be interpreted independently or jointly by several neural networks. The sign itself can be recognized as a traffic sign by a deployed first neural network (e.g., a trained neural network), and the text "Flashing lights indicate icy conditions" can be interpreted by a deployed second neural network that informs the vehicle's path planning software (preferably executing on a CPU complex) that icy conditions exist when the flashing lights are detected. The flashing lights can be identified by a deployed third neural network operating across multiple frames, informing the vehicle's path planning software of the presence (or absence) of the flashing lights. All three neural networks can operate simultaneously, for example, within a DLA and / or on a GPU 1108.

[0161] In some examples, the CNN used for facial recognition and owner identification can use data from camera sensors to identify the presence of an authorized driver and / or owner of vehicle 1100. A processing engine always on the sensors can be used to unlock the vehicle and turn on the lights when the owner approaches the driver's door, and in safe mode, to disable the vehicle when the owner leaves. In this way, SoC 1104 provides security against theft and / or carjacking.

[0162] In another example, the CNN used for emergency vehicle detection and identification can use data from microphone 1196 to detect and identify emergency vehicle siren. In contrast to conventional systems that use a general classifier to detect siren and manually extract features, SoC 1104 uses a CNN to classify environmental and urban sounds as well as visual data. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative shut-off rate of emergency vehicles (e.g., by using the Doppler effect). The CNN can also be trained to identify emergency vehicles specific to the localized area in which the vehicle operates, as identified by GNSS sensor 1158. Thus, for example, when operating in Europe, the CNN will seek to detect European siren, and when operating in the United States, the CNN will seek to identify siren only in North America. Once an emergency vehicle is detected, with the assistance of ultrasonic sensor 1162, the control program can be used to execute emergency vehicle safety routines, causing the vehicle to slow down, pull over to the side of the road, stop, and / or idle until the emergency vehicle passes.

[0163] The vehicle may include a CPU 1118 (e.g., a discrete CPU or dCPU) that can be coupled to the SoC 1104 via a high-speed interconnect (e.g., PCIe). The CPU 1118 may include, for example, an x86 processor. The CPU 1118 can be used to perform any of a wide variety of functions, including, for example, arbitrating the results of potential inconsistencies between ADAS sensors and the SoC 1104, and / or monitoring the status and health of the controller 1136 and / or the infotainment SoC 1130.

[0164] Vehicle 1100 may include a GPU 1120 (e.g., a discrete GPU or dGPU) that can be coupled to SoC 1104 via a high-speed interconnect (e.g., NVIDIA's NVLINK). GPU 1120 may provide additional artificial intelligence capabilities, for example by executing redundant and / or different neural networks, and can be used to train and / or update neural networks based at least in part on inputs (e.g., sensor data) from sensors of vehicle 1100.

[0165] Vehicle 1100 may also include a network interface 1124, which may include one or more wireless antennas 1126 (e.g., one or more wireless antennas for different communication protocols, such as cellular antennas, Bluetooth antennas, etc.). Network interface 1124 can be used to enable wireless connectivity via the Internet to the cloud (e.g., with server 1178 and / or other network devices), with other vehicles, and / or with computing devices (e.g., passenger client devices). For communication with other vehicles, a direct link can be established between the two vehicles, and / or an indirect link can be established (e.g., across networks and via the Internet). A direct link can be provided using a vehicle-to-vehicle communication link. The vehicle-to-vehicle communication link can provide vehicle 1100 with information about vehicles approaching vehicle 1100 (e.g., vehicles in front, to the side, and / or behind vehicle 1100). This functionality may be part of vehicle 1100's cooperative adaptive cruise control function.

[0166] Network interface 1124 may include a SoC that provides modulation and demodulation functions and enables controller 1136 to communicate via a wireless network. Network interface 1124 may include an RF front-end for up-conversion from baseband to RF and down-conversion from RF to baseband. Frequency conversion can be performed using known processes and / or using a superheterodyne process. In some examples, the RF front-end functionality may be provided by a separate chip. The network interface may include wireless functions for communication via LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-Wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0167] Vehicle 1100 may also include data storage 1128, which may include off-chip (e.g., off-chip SoC 1104) storage devices. Data storage 1128 may include one or more storage elements, including RAM, SRAM, DRAM, VRAM, flash memory, hard disk, and / or other components and / or devices capable of storing at least one bit of data.

[0168] Vehicle 1100 may also include a GNSS sensor 1158. The GNSS sensor 1158 (e.g., GPS, assisted GPS sensor, differential GPS (DGPS) sensor, etc.) is used for auxiliary mapping, sensing, occupancy grid generation, and / or path planning functions. Any number of GNSS sensors 1158 can be used, including, for example, but not limited to, GPS using a USB connector with an Ethernet-to-serial (RS-232) bridge.

[0169] Vehicle 1100 may also include a RADAR sensor 1160. The RADAR sensor 1160 can be used by vehicle 1100 for remote vehicle detection even in dark and / or inclement weather conditions. The RADAR functional safety level may be ASIL B. The RADAR sensor 1160 can use CAN and / or bus 1102 (e.g., to transmit data generated by the RADAR sensor 1160) for control and access to object tracking data, and in some examples, Ethernet access for accessing raw data. A wide variety of RADAR sensor types can be used. For example, and without limitation, the RADAR sensor 1160 can be adapted for front, rear, and side RADAR use. In some examples, a pulse Doppler RADAR sensor is used.

[0170] RADAR sensor 1160 can include different configurations, such as long-range with a narrow field of view, short-range with a wide field of view, short-range side coverage, etc. In some examples, long-range RADAR can be used for adaptive cruise control functions. A long-range RADAR system can provide a wide field of view (e.g., within 250m) achieved through two or more independent scans. RADAR sensor 1160 can help distinguish between stationary and moving objects and can be used by ADAS systems for emergency braking assist and forward collision warning. Long-range RADAR sensors can include a single-site multi-mode RADAR with multiple (e.g., six or more) fixed RADAR antennas and high-speed CAN and FlexRay interfaces. In an example with six antennas, the four central antennas can create a focused beam pattern designed to record the vehicle 1100's surroundings at higher rates with minimal traffic interference from adjacent lanes. The other two antennas can extend the field of view, enabling rapid detection of vehicles entering or leaving the vehicle 1100's lane.

[0171] As an example, a mid-range RADAR system can include a range of up to 1160m (front) or 80m (rear) and a field of view of up to 42 degrees (front) or 1150 degrees (rear). Short-range RADAR systems can include, but are not limited to, RADAR sensors designed to be mounted at both ends of the rear bumper. When mounted at both ends of the rear bumper, such a RADAR sensor system can create two beams that continuously monitor blind spots behind and beside the vehicle.

[0172] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0173] Vehicle 1100 may also include ultrasonic sensors 1162. Ultrasonic sensors 1162, which may be positioned at the front, rear, and / or sides of vehicle 1100, can be used for parking assistance and / or creating and updating occupancy grids. A wide variety of ultrasonic sensors 1162 can be used, and different ultrasonic sensors 1162 can be used for different detection ranges (e.g., 2.5m, 4m). Ultrasonic sensors 1162 can operate at functional safety level ASIL B.

[0174] Vehicle 1100 may include a LIDAR sensor 1164. The LIDAR sensor 1164 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LIDAR sensor 1164 may be of functional safety level ASIL B. In some examples, vehicle 1100 may include multiple LIDAR sensors 1164 (e.g., two, four, six, etc.) that can use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0175] In some examples, the LiDAR sensor 1164 may be able to provide a list of objects and their distances within a 360-degree field of view. Commercially available LiDAR sensors 1164 may have an advertising range of, for example, approximately 1100m, with an accuracy of 2cm-3cm, and support for 1100Mbps Ethernet connectivity. In some examples, one or more non-protruding LiDAR sensors 1164 may be used. In such examples, the LiDAR sensor 1164 may be implemented as a small device that can be embedded in the front, rear, sides, and / or corners of the vehicle 1100. In such examples, the LiDAR sensor 1164 may provide a horizontal field of view of up to 120 degrees and a vertical field of view of 35 degrees, even for low-reflectivity objects, with a range of 200m. Front-mounted LiDAR sensors 1164 may be configured for a horizontal field of view between 45 and 135 degrees.

[0176] In some examples, LiDAR technologies such as 3D flash LiDAR can also be used. 3D flash LiDAR uses a flash of laser light as the emission source to illuminate the vehicle's surroundings up to approximately 200 meters. A flash LiDAR unit includes a receiver that records the laser pulse propagation time and reflected light on each pixel, which in turn corresponds to the range from the vehicle to the object. Flash LiDAR allows for the generation of highly accurate and distortion-free images of the surrounding environment using each laser flash. In some examples, four flash LiDAR sensors can be deployed, one on each side of the vehicle. Available 3D flash LiDAR systems include solid-state 3D staring array LiDAR cameras (e.g., non-scanning LiDAR devices) without moving parts other than a fan. Flash LiDAR devices can use 5 nanosecond Class I (eye-safe) laser pulses per frame and can capture reflected laser light in the form of a 3D range point cloud and co-registered intensity data. By using a flash LIDAR, and because a flash LIDAR is a solid-state device with no moving parts, the LIDAR sensor 1164 is less susceptible to motion blur, vibration, and / or shock.

[0177] The vehicle may also include an IMU sensor 1166. In some examples, the IMU sensor 1166 may be located at the center of the rear axle of the vehicle 1100. The IMU sensor 1166 may include, for example, but not limited to, an accelerometer, a magnetometer, a gyroscope, a magnetic compass, and / or other sensor types. In some examples, such as in a six-axis application, the IMU sensor 1166 may include an accelerometer and a gyroscope, while in a nine-axis application, the IMU sensor 1166 may include an accelerometer, a gyroscope, and a magnetometer.

[0178] In some embodiments, the IMU sensor 1166 can be implemented as a miniature, high-performance GPS-assisted inertial navigation system (GPS / INS) that combines a microelectromechanical system (MEMS) inertial sensor, a high-sensitivity GPS receiver, and an advanced Kalman filter algorithm to provide estimates of position, velocity, and attitude. Thus, in some examples, the IMU sensor 1166 can enable the vehicle 1100 to estimate heading without input from a magnetic sensor by directly observing and correlating velocity changes from GPS to the IMU sensor 1166. In some examples, the IMU sensor 1166 and the GNSS sensor 1158 can be combined into a single integrated unit.

[0179] The vehicle may include a microphone 1196 placed in and / or around the vehicle 1100. Among other things, the microphone 1196 may be used for emergency vehicle detection and identification.

[0180] The vehicle may also include any number of camera types, including stereo camera 1168, wide-angle camera 1170, infrared camera 1172, surround camera 1174, long-range and / or mid-range camera 1198, and / or other camera types. These cameras can be used to capture image data around the entire perimeter of the vehicle 1100. The types of cameras used depend on the embodiment and the requirements of the vehicle 1100, and any combination of camera types can be used to provide the necessary coverage around the vehicle 1100. Furthermore, the number of cameras may vary depending on the embodiment. For example, the vehicle may include six cameras, seven cameras, ten cameras, twelve cameras, and / or another number of cameras. As an example and without limitation, these cameras may support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each of the cameras is described herein with respect to... Figure 11A and Figure 11B It was described in more detail.

[0181] Vehicle 1100 may also include vibration sensor 1142. Vibration sensor 1142 can measure vibrations of vehicle components such as axles. For example, changes in vibration can indicate changes in the road surface. In another example, when two or more vibration sensors 1142 are used, differences between vibrations can be used to determine friction or slippage on the road surface (e.g., when there is a vibration difference between a power drive shaft and a free-rotating shaft).

[0182] Vehicle 1100 may include ADAS system 1138. In some examples, ADAS system 1138 may include SoC. ADAS system 1138 may include autonomous / adaptive / automatic cruise control (ACC), cooperative adaptive cruise control (CACC), forward collision warning (FCW), automatic emergency braking (AEB), lane departure warning (LDW), lane keeping assist (LKA), blind spot warning (BSW), rear cross traffic warning (RCTW), collision warning system (CWS), lane centering (LC) and / or other features and functions.

[0183] The ACC system can use RADAR sensor 1160, LIDAR sensor 1164, and / or a camera. The ACC system may include longitudinal ACC and / or lateral ACC. Longitudinal ACC monitors and controls the distance to vehicles immediately in front of vehicle 1100 and automatically adjusts the vehicle speed to maintain a safe distance. Lateral ACC performs distance holding and, if necessary, advises vehicle 1100 to change lanes. Lateral ACC is associated with other ADAS applications such as LCA and CWS.

[0184] CACC uses information from other vehicles, which can be received indirectly from other vehicles via a wireless link or through a network connection (e.g., via the Internet) through network interface 1124 and / or wireless antenna 1126. Direct links can be provided by vehicle-to-vehicle (V2V) communication links, while indirect links can be infrastructure-to-vehicle (I2V) communication links. Typically, the V2V communication concept provides information about vehicles immediately ahead (e.g., vehicles immediately in front of vehicle 1100 and in the same lane), while the I2V communication concept provides information about traffic further ahead. A CACC system can include either or both of these I2V and V2V information sources. Given information about vehicles ahead of vehicle 1100, CACC can be more reliable, and it has the potential to improve traffic flow and reduce road congestion.

[0185] The Forward-Looking Warning (FCW) system is designed to alert the driver to hazards, enabling the driver to take corrective action. The FCW system uses a front-facing camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker, and / or vibrating components. The FCW system can provide warnings in the form of, for example, audible, visual, haptic, and / or rapid braking pulses.

[0186] The AEB system detects an impending forward collision with another vehicle or other object and can automatically apply the brakes if the driver does not take corrective action within a specified time or distance parameter. The AEB system can use a front-facing camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC. When the AEB system detects a hazard, it typically first alerts the driver to take corrective action to avoid a collision, and if the driver does not take corrective action, the AEB system can automatically apply the brakes to attempt to prevent or at least mitigate the effects of the predicted collision. The AEB system may include technologies such as dynamic brake support and / or collision proximity braking.

[0187] The Lane Departure Warning (LDW) system provides visual, auditory, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver when the vehicle crosses lane markings. When the driver indicates intentional lane departure, the LDW system is deactivated by activating a turn signal. The LDW system can utilize a front-facing camera coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker, and / or vibrating components.

[0188] The LKA system is a variant of the LDW system. If vehicle 1100 begins to leave the lane, the LKA system provides steering input or braking to correct vehicle 1100.

[0189] The BSW system detects and warns the driver of vehicles in the vehicle's blind spot. The BSW system can provide visual, auditory, and / or tactile alerts to indicate that merging or changing lanes is unsafe. The system can provide additional warnings when the driver uses turn signals. The BSW system can use a rear-facing camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker, and / or vibrating components.

[0190] The RCTW system can provide visual, auditory, and / or tactile notifications when an object is detected outside the range of the rear camera while the vehicle 1100 is reversing. Some RCTW systems include AEB to ensure the application of the vehicle's brakes to avoid a collision. The RCTW system may use one or more rear-mounted RADAR sensors 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker, and / or vibrating components.

[0191] Conventional ADAS systems can be prone to false positives, which can be annoying and distracting for the driver, but typically not catastrophic, as they alert the driver and allow them to determine whether a safe condition truly exists and take appropriate action. However, in the autonomous vehicle 1100, in the event of conflicting results, the vehicle 1100 itself must decide whether to heed the results from the main computer or auxiliary computer (e.g., the first controller 1136 or the second controller 1136). For example, in some embodiments, the ADAS system 1138 may be a backup and / or auxiliary computer for providing perception information to a backup computer rationality module. The backup computer rationality monitor may run redundant and varied software on hardware components to detect faults in perception and dynamic driving tasks. Outputs from the ADAS system 1138 may be provided to a supervisory MCU. If the outputs from the main computer and the auxiliary computer conflict, the supervisory MCU must determine how to reconcile the conflict to ensure safe operation.

[0192] In some examples, the master computer can be configured to provide a confidence score to the supervisory MCU, indicating the master computer's confidence level in the selected result. If the confidence score exceeds a threshold, the supervisory MCU can follow the master computer's direction regardless of whether the auxiliary computer provides conflicting or inconsistent results. If the confidence score does not meet the threshold and the master and auxiliary computers indicate different results (e.g., conflict), the supervisory MCU can arbitrate between these computers to determine the appropriate result.

[0193] The supervisory MCU can be configured to run a neural network trained and configured to determine the conditions under which the auxiliary computer provides a false alarm, based at least in part on outputs from both the host and auxiliary computers. Thus, the neural network in the supervisory MCU can learn when the output of the auxiliary computer can be trusted and when it cannot. For example, when the auxiliary computer is a RADAR-based FCW system, the neural network in the supervisory MCU can learn when the FCW system is identifying a metallic object that is not actually dangerous, such as a drain grid or manhole cover that triggers an alarm. Similarly, when the auxiliary computer is a camera-based LDW system, the neural network in the supervisory MCU can learn to ignore the LDW when a cyclist or pedestrian is present and lane departure is actually the safest strategy. In embodiments that include a neural network running on the supervisory MCU, the supervisory MCU may include at least one of a DLA or GPU suitable for running the neural network using associated memory. In a preferred embodiment, the supervisory MCU may include a component of SoC 1104 and / or be included as a component of SoC 1104.

[0194] In other examples, ADAS system 1138 may include an auxiliary computer that performs ADAS functions using conventional computer vision rules. This allows the auxiliary computer to use classic computer vision rules (if-then), and the presence of neural networks in the supervising MCU can improve reliability, safety, and performance. For example, diverse implementations and intentional non-identity make the entire system more fault-tolerant, especially for failures caused by software (or software-hardware interface) functionality. For instance, if a software vulnerability or bug exists in the software running on the host computer and non-identical software code running on the auxiliary computer provides the same overall result, the supervising MCU can be more confident that the overall result is correct and that the vulnerability in the software or hardware on the host computer does not cause a substantial error.

[0195] In some examples, the output of ADAS system 1138 can be fed to the perception block and / or the dynamic driving task block of the main computer. For example, if ADAS system 1138 issues a forward collision warning because an object is immediately in front, the perception block can use this information when identifying the object. In other examples, the assistance computer can have its own neural network, which is trained and thus reduces the risk of false positives as described herein.

[0196] Vehicle 1100 may also include an infotainment SoC 1130 (e.g., an in-vehicle infotainment system (IVI)). Although illustrated and described as an SoC, the infotainment system may not be an SoC and may include two or more discrete components. The infotainment SoC 1130 may include a combination of hardware and software that can be used to provide vehicle 1100 with audio (e.g., music, personal digital assistant, navigation instructions, news, radio, etc.), video (e.g., TV, movies, streaming media, etc.), telephone (e.g., hands-free calling), network connectivity (e.g., LTE, WiFi, etc.) and / or information services (e.g., navigation system, rear parking assistance, radio data system, vehicle-related information such as fuel level, total coverage distance, brake fuel level, fuel level, door opening / closing, air filter information, etc.). For example, the infotainment SoC 1130 may include a radio, disc player, navigation system, video player, USB and Bluetooth connectivity, in-vehicle computer, in-vehicle entertainment, WiFi, steering wheel audio controls, hands-free voice controls, head-up display (HUD), HMI display 1134, telematics device, control panel (e.g., for controlling and / or interacting with various components, features, and / or systems) and / or other components. The infotainment SoC 1130 may further be used to provide information (e.g., visual and / or auditory) to users of the vehicle, such as information from the ADAS system 1138, autonomous driving information such as planned vehicle maneuvers, trajectories, surrounding environment information (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0197] The infotainment SoC 1130 may include GPU functionality. The infotainment SoC 1130 can communicate with other devices, systems, and / or components of the vehicle 1100 via bus 1102 (e.g., CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 1130 may be coupled to a supervisory MCU, allowing the GPU of the infotainment system to perform some autonomous driving functions in the event of a failure of the main controller 1136 (e.g., the primary and / or backup computer of the vehicle 1100). In such an example, the infotainment SoC 1130 may place the vehicle 1100 into a driver-safe parking mode as described herein.

[0198] Vehicle 1100 may also include instrument cluster 1132 (e.g., digital instrument panel, electronic instrument cluster, digital instrument panel, etc.). Instrument cluster 1132 may include a controller and / or supercomputer (e.g., a discrete controller or supercomputer). Instrument cluster 1132 may include a set of instruments such as speedometer, fuel level, oil pressure, tachometer, odometer, turn indicator, shift position indicator, seatbelt warning light, parking brake warning light, engine malfunction indicator, airbag (SRS) system information, lighting controls, safety system controls, navigation information, etc. In some examples, information may be displayed and / or shared between infotainment SoC 1130 and instrument cluster 1132. In other words, instrument cluster 1132 may be included as part of infotainment SoC 1130, or vice versa.

[0199] Figure 11D For cloud-based servers and according to some embodiments of this disclosure Figure 11A The following is a system diagram illustrating communication between example autonomous vehicles 1100. System 1176 may include server 1178, network 1190, and vehicles including vehicle 1100. Server 1178 may include multiple GPUs 1184(A)-1284(H) (collectively referred to herein as GPU 1184), PCIe switches 1182(A)-1182(H) (collectively referred to herein as PCIe switch 1182), and / or CPUs 1180(A)-1180(B) (collectively referred to herein as CPU 1180). GPU 1184, CPU 1180, and PCIe switches may be interconnected with high-speed interconnects and / or PCIe connections 1186, such as, but not limited to, NVLink interface 1188 developed by NVIDIA. In some examples, GPU 1184 is connected via NVLink and / or NVSwitch SoC, and GPU 1184 and PCIe switch 1182 are connected via PCIe interconnect. Although the diagram illustrates eight GPUs 1184, two CPUs 1180, and two PCIe switches, it is not intended to be limiting. Depending on the embodiment, each of the servers 1178 may include any number of GPUs 1184, CPUs 1180, and / or PCIe switches. For example, each of the servers 1178 may include eight, sixteen, thirty-two, and / or more GPUs 1184.

[0200] Server 1178 can receive image data from vehicles via network 1190, representing images of unexpected or altered road conditions such as recently commenced roadworks. Server 1178 can also transmit neural network 1192, updated neural network 1192, and / or map information 1194, including information about traffic and road conditions, to vehicles via network 1190. Updates to map information 1194 may include updates to HD map 1122, such as information about construction sites, potholes, bends, floods, or other obstacles. In some examples, neural network 1192, updated neural network 1192, and / or map information 1194 may have been generated from new training and / or data received from any number of vehicles in the environment, and / or based on experience gained from training performed at a data center (e.g., using server 1178 and / or other servers).

[0201] Server 1178 can be used to train machine learning models (e.g., neural networks) based on training data. Training data can be generated by the vehicle and / or generated in a simulation (e.g., using a game engine). In some examples, the training data is labeled (e.g., where the neural network benefits from supervised learning) and / or undergoes other preprocessing, while in other examples, the training data is not labeled and / or preprocessed (e.g., where the neural network does not require supervised learning). Training can be performed according to any one or more classes of machine learning techniques, including but not limited to: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, joint learning, transfer learning, feature learning (including principal component and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including alternative dictionary learning), rule-based machine learning, anomaly detection, and any variations or combinations thereof. Once the machine learning model is trained, it can be used by the vehicle (e.g., transmitted to the vehicle via network 1190), and / or the machine learning model can be used by server 1178 to remotely monitor the vehicle.

[0202] In some examples, server 1178 can receive data from a vehicle and apply that data to a state-of-the-art real-time neural network for real-time intelligent inference. Server 1178 may include a deep learning supercomputer powered by GPU 1184 and / or a dedicated AI computer, such as the DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 1178 may include a deep learning infrastructure in a data center that uses only CPU power.

[0203] The deep learning infrastructure of server 1178 may be capable of rapid real-time inference and can be used to assess and verify the health status of the processor, software, and / or associated hardware in vehicle 1100. For example, the deep learning infrastructure may receive periodic updates from vehicle 1100, such as image sequences and / or objects located in those image sequences that vehicle 1100 has already located (e.g., via computer vision and / or other machine learning object classification techniques). The deep learning infrastructure may run its own neural network to identify objects and compare them with objects identified by vehicle 1100. If the results do not match and the infrastructure concludes that the AI ​​in vehicle 1100 has malfunctioned, then server 1178 may transmit a signal to vehicle 1100 instructing the vehicle's fail-safe computer to take control, notify passengers, and complete a safe stopping operation.

[0204] For inference, server 1178 may include GPU 1184 and one or more programmable inference accelerators (such as NVIDIA's TensorRT). The combination of a GPU-powered server and inference acceleration enables real-time response. In other examples, such as where performance is less critical, CPU, FPGA, and other processor-powered servers can be used for inference.

[0205] Example computing device

[0206] Figure 12 This is a block diagram of an example computing device 1200 suitable for implementing some embodiments of the present disclosure. The computing device 1200 may include an interconnect system 1202 directly or indirectly coupled to the following devices: a memory 1204, one or more central processing units (CPUs) 1206, one or more graphics processing units (GPUs) 1208, a communication interface 1210, input / output (I / O) ports 1212, input / output components 1214, a power supply 1216, one or more presentation components 1218 (e.g., one or more displays), and one or more logic units 1220. In at least one embodiment, the computing device 1200 may include one or more virtual machines (VMs), and / or any component thereof may include virtual components (e.g., virtual hardware components). For a non-limiting example, one or more GPUs 1208 may include one or more vGPUs, one or more CPUs 1206 may include one or more vCPUs, and / or one or more logic units 1220 may include one or more virtual logic units. Thus, (one or more) computing devices 1200 may include discrete components (e.g., a full GPU dedicated to computing device 1200), virtual components (e.g., a portion of the GPU dedicated to computing device 1200), or a combination thereof.

[0207] although Figure 12 The various blocks are shown as connected via interconnect system 1202 using lines, but this is not intended to be limiting and is merely for clarity. For example, in some embodiments, presentation component 1218 (such as a display device) may be considered I / O component 1214 (e.g., if the display is a touchscreen). As another example, CPU 1206 and / or GPU 1208 may include memory (e.g., memory 1204 may represent a storage device other than the memory of GPU 1208, CPU 1206, and / or other components). In other words, Figure 12 The computing devices described are for illustrative purposes only. No distinction is made between such categories as “workstation,” “server,” “laptop computer,” “desktop computer,” “tablet computer,” “client device,” “mobile device,” “handheld device,” “game console,” “electronic control unit (ECU),” “virtual reality system,” and / or other device or system types, as all are considered within the scope of… Figure 12 Within the scope of computing devices.

[0208] Interconnect system 1202 may represent one or more links or buses, such as address buses, data buses, control buses, or combinations thereof. Interconnect system 1202 may include one or more bus or link types, such as Industry Standard Architecture (ISA) bus, Extended Industry Standard Architecture (EISA) bus, Video Electronics Standards Association (VESA) bus, Peripheral Component Interconnect (PCI) bus, Fast Peripheral Component Interconnect (PCIe) bus, and / or another type of bus or link. In some embodiments, there is a direct connection between components. As an example, CPU 1206 may be directly connected to memory 1204. Further, CPU 1206 may be directly connected to GPU 1208. In cases where there is a direct or point-to-point connection between components, interconnect system 1202 may include a PCIe link to perform the connection. In these examples, a PCI bus is not required to be included in computing device 1200.

[0209] The memory 1204 may include any computer-readable medium from a variety of computer-readable media. The computer-readable medium may be any available medium accessible by the computing device 1200. The computer-readable medium may include volatile and non-volatile media, as well as removable and non-removable media. By way of example and not limitation, the computer-readable medium may include computer storage media and communication media.

[0210] Computer storage media may include volatile and non-volatile media and / or removable and non-removable media implemented with any method or technology for storing information such as computer-readable instructions, data structures, program modules and / or other data types. For example, memory 1204 may store computer-readable instructions (e.g., representing (one or more) programs and / or (one or more) program elements, such as an operating system). Computer storage media may include, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic tape cassettes, magnetic tape, disk storage devices or other magnetic storage devices, or any other medium that can be used to store the desired information and is accessible by computing device 1200. As used herein, computer storage media does not include the signal itself.

[0211] Computer storage media can embody computer-readable instructions, data structures, program modules, and / or other data types in modulated data signals such as carrier waves or other transmission mechanisms, and includes any information transmission medium. The term "modulated data signal" can refer to a signal whose one or more characteristics are set or altered in a manner that encodes information in the signal. By way of example and not limitation, computer storage media can include wired media (such as wired networks or direct wired connections) and wireless media (such as acoustic, RF, infrared, and other wireless media). Any combination of the above should also be included within the scope of computer-readable media.

[0212] CPU 1206 may be configured to execute at least some of computer-readable instructions to control one or more components of computing device 1200 to perform one or more of the methods and / or processes described herein. Each CPU 1206 may contain one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of handling numerous software threads simultaneously. CPU 1206 may contain any type of processor and may contain different types of processors depending on the type of computing device 1200 implemented (e.g., processors with fewer cores for mobile devices and processors with more cores for servers). For example, depending on the type of computing device 1200, the processor may be an advanced RISC machine (ARM) processor implemented using Reduced Instruction Set Computing (RISC) or an x86 processor implemented using Complex Instruction Set Computing (CISC). In addition to one or more microprocessors or supplementary coprocessors (such as math coprocessors), computing device 1200 may also include one or more CPUs 1206.

[0213] In addition to or in lieu of one or more CPUs 1206, one or more GPUs 1208 may be configured to execute at least some of computer-readable instructions to control one or more components of computing device 1200 to perform one or more of the methods and / or processes described herein. One or more GPUs 1208 may be integrated GPUs (e.g., with one or more CPUs 1206) and / or one or more GPUs 1208 may be discrete GPUs. In embodiments, one or more GPUs 1208 may be coprocessors of one or more CPUs 1206. GPUs 1208 may be used by computing device 1200 to render graphics (e.g., 3D graphics) or perform general-purpose computing. For example, GPUs 1208 may be used for general-purpose computing on a GPU (GPGPU). GPUs 1208 may contain hundreds or thousands of cores capable of handling hundreds or thousands of software threads simultaneously. GPU 1208 can generate pixel data for an output image in response to rendering commands (e.g., rendering commands received from CPU 1206 via a host interface). GPU 1208 may include graphics memory (e.g., display memory) for storing pixel data or any other suitable data (e.g., GPGPU data). Display memory may be included as part of memory 1204. GPU 1208 may include two or more GPUs operating in parallel (e.g., via links). The links may directly connect the GPUs (e.g., using NVLINK) or may connect the GPUs via a switch (e.g., using NVSwitch). When combined, each GPU 1208 may generate pixel data or GPGPU data for different portions of the output or for different outputs (e.g., a first GPU for a first image and a second GPU for a second image). Each GPU may include its own memory or may share memory with other GPUs.

[0214] In addition to or in lieu of CPU 1206 and / or GPU 1208, logic unit 1220 may be configured to execute at least some of computer-readable instructions to control one or more components of computing device 1200 to perform one or more of the methods and / or processes described herein. In embodiments, one or more CPUs 1206, one or more GPUs 1208, and / or one or more logic units 1220 may perform any combination of methods, processes, and / or portions thereof discretely or jointly. One or more logic units 1220 may be a portion of one or more CPUs 1206 and / or GPUs 1208 and / or integrated into one or more CPUs 1206 and / or GPUs 1208, and / or one or more logic units 1220 may be discrete components or otherwise external to CPUs 1206 and / or GPUs 1208. In an embodiment, one or more of the logic units 1220 may be coprocessors of one or more of the CPU 1206 and / or one or more of the GPU 1208.

[0215] Examples of logic unit 1220 include one or more processing cores and / or components thereof, such as data processing unit (DPU), tensor core (TC), tensor processing unit (TPU), pixel vision core (PVC), vision processing unit (VPU), graphics processing cluster (GPC), texture processing cluster (TPC), streaming multiprocessor (SM), tree lateral unit (TTU), artificial intelligence accelerator (AIA), deep learning accelerator (DLA), arithmetic logic unit (ALU), application-specific integrated circuit (ASIC), floating-point unit (FPU), input / output (I / O) element, peripheral component interconnect (PCI) or fast peripheral component interconnect (PCIe) element, etc.

[0216] Communication interface 1210 may include one or more receivers, transmitters, and / or transceivers enabling computing device 1200 to communicate with other computing devices via electronic communication networks (including wired and / or wireless communications). Communication interface 1210 may include components and functions for enabling communication over any of a plurality of different networks, such as wireless networks (e.g., Wi-Fi, Z-Wave, Bluetooth, Bluetooth LE, ZigBee, etc.), wired networks (e.g., via Ethernet or wirelessband communication), low-power wide area networks (e.g., LoRaWAN, SigFox, etc.), and / or the Internet. In one or more embodiments, one or more logic units 1220 and / or communication interface 1210 may include one or more data processing units (DPUs) for directly transmitting data received via a network and / or via interconnect system 1202 to one or more GPUs 1208 (e.g., their memory).

[0217] I / O port 1212 enables computing device 1200 to be logically coupled to other devices including I / O component 1214, (one or more) presentation component 1218, and / or other components, some of which may be built into (e.g., integrated into) computing device 1200. Illustrative I / O component 1214 includes microphones, mice, keyboards, joysticks, gamepads, game controllers, satellite dish antennas, scanners, printers, wireless devices, etc. I / O component 1214 can provide a natural user interface (NUI) that processes aerial gestures, voice, or other physiological input generated by the user. In some cases, input may be transmitted to appropriate network elements for further processing. The NUI can implement any combination of voice recognition, pen recognition, facial recognition, biometric recognition, on-screen and near-screen gesture recognition, aerial gestures, head and eye tracking, and touch recognition (as described in more detail below) associated with the display of computing device 1200. Computing device 1200 may include depth cameras for gesture detection and recognition, such as stereo camera systems, infrared camera systems, RGB camera systems, touchscreen technology, and combinations thereof. Additionally, the computing device 1200 may include an accelerometer or gyroscope (e.g., as part of an inertial measurement unit (IMU)) that enables motion detection. In some examples, the computing device 1200 may use the output of the accelerometer or gyroscope to render immersive augmented reality or virtual reality.

[0218] Power supply 1216 may include a hardwired power supply, a battery power supply, or a combination thereof. Power supply 1216 may provide power to computing device 1200 to enable the components of computing device 1200 to operate.

[0219] The presentation component 1218 may include a display (e.g., a monitor, touchscreen, television screen, head-up display (HUD), other display types, or combinations thereof), speakers, and / or other presentation components. The presentation component 1218 may receive data from other components (e.g., GPU 1208, CPU 1206, DPU, etc.) and output the data (e.g., as images, videos, sounds, etc.).

[0220] Example Data Center

[0221] Figure 13 An example data center 1300 that may be used in at least one embodiment of this disclosure is shown. The data center 1300 may include a data center infrastructure layer 1310, a framework layer 1320, a software layer 1330, and / or an application layer 1340.

[0222] like Figure 13 As shown, the data center infrastructure layer 1310 may include a resource coordinator 1312, grouped computing resources 1314, and node computing resources (“nodes CRs”) 1316(1)-1316(N), where “N” represents any complete positive integer. In at least one embodiment, nodes CRs 1316(1)-1316(N) may include, but are not limited to, any number of central processing units (CPUs) or other processors (including DPUs, accelerators, field-programmable gate arrays (FPGAs), graphics processing units or graphics processing units (GPUs), etc.), memory devices (e.g., dynamic read-only memory), storage devices (e.g., solid-state or disk drives), network input / output (NW I / O) devices, network switches, virtual machines (VMs), power modules and / or cooling modules, etc. In some embodiments, one or more node CRs from nodes CRs 1316(1)-1316(N) may correspond to servers having one or more of the aforementioned computing resources. In addition, in some embodiments, nodes CRs1316(1)-13161(N) may include one or more virtual components, such as vGPU, vCPU, etc., and / or one or more nodes CRs1316(1)-1316(N) may correspond to virtual machines (VMs).

[0223] In at least one embodiment, the grouped computing resources 1314 may include individual groups of node CRs 1316 housed within one or more racks (not shown), or a plurality of racks housed within a data center in different geographical locations (also not shown). Individual groups of node CRs 1316 within the grouped computing resources 1314 may include grouped computing, networking, memory, or storage resources that can be configured or allocated to support one or more workloads. In at least one embodiment, a plurality of node CRs 1316, including CPUs, GPUs, DPUs, and / or other processors, may be grouped within one or more racks to provide computing resources to support one or more workloads. One or more racks may also include any number of power modules, cooling modules, and / or network switches in any combination.

[0224] Resource coordinator 1312 may be configured or otherwise control one or more nodes CRs 1316(1)-1316(N) and / or grouped computing resources 1314. In at least one embodiment, resource coordinator 1312 may include a Software Design Infrastructure (SDI) management entity for data center 1300. Resource coordinator 1312 may include hardware, software, or some combination thereof.

[0225] In at least one embodiment, such as Figure 13 As shown, framework layer 1320 may include a job scheduler 1333, a configuration manager 1334, a resource manager 1336, and / or a distributed file system 1338. Framework layer 1320 may include a framework for software 1332 supporting software layer 1330 and / or one or more applications 1342 supporting application layer 1340. Software 1332 or application 1342 may respectively contain web-based service software or applications, such as those provided by Amazon Web Services, Google Cloud, and Microsoft Azure. Framework layer 1320 may be, but is not limited to, free and open-source software web application frameworks (such as Apache Spark) that can utilize distributed file system 1338 for large-scale data processing (e.g., "big data"). TM(Hereinafter referred to as "Spark") is a type of resource manager. In at least one embodiment, the job scheduler 1333 may include Spark drivers to facilitate the scheduling of workloads supported by different layers of data center 1300. The configuration manager 1334 may be able to configure different layers, such as software layer 1330 and framework layer 1320 (which includes Spark and distributed file system 1338 for supporting large-scale data processing). The resource manager 1336 may be able to manage compute resources mapped to or allocated to clusters of distributed file system 1338 and job scheduler 1333 or to support clusters of distributed file system 1338 and job scheduler 1333. In at least one embodiment, the clustered or grouped compute resources may include grouped compute resources 1314 in data center infrastructure layer 1310. The resource manager 1336 may coordinate with resource coordinator 1312 to manage these mapped or allocated compute resources.

[0226] In at least one embodiment, the software 1332 included in software layer 1330 may include software used in at least a portion of the nodes CRs 1316(1)-1316(N), the grouped computing resources 1314, and / or the distributed file system 1338 of framework layer 1320. One or more types of software may include, but are not limited to, internet web search software, email virus scanning software, database software, and streaming video content software.

[0227] In at least one embodiment, the application 1342 included in the application layer 1340 may include one or more types of applications used at least in part by nodes CRs 1316(1)-1316(N), grouped computing resources 1314, and / or the distributed file system 1338 of the framework layer 1320. 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.), and / or other machine learning applications used in combination with one or more embodiments.

[0228] In at least one embodiment, any of the configuration manager 1334, resource manager 1336, and resource coordinator 1312 can implement any number and type of self-modification actions based on any amount and type of data acquired in any technically feasible manner. Self-modification actions can free data center operators of data center 1300 from making potentially poor configuration decisions and may prevent underutilization and / or poor performance of the data center.

[0229] According to one or more embodiments described herein, data center 1300 may include tools, services, software, or other resources to train one or more machine learning models or to use one or more machine learning models to predict or infer information. For example, one or more machine learning models can be trained by using the software and / or computing resources described above with respect to data center 1300 to compute weight parameters according to a neural network architecture. In at least one embodiment, a trained or deployed machine learning model corresponding to one or more neural networks can be used to infer or predict information using the resources described above with respect to data center 1300 by using weight parameters computed through one or more training techniques (such as, but not limited to, those described herein).

[0230] In at least one embodiment, the data center 1300 may use a CPU, application-specific integrated circuit (ASIC), GPU, FPGA, and / or other hardware (or corresponding virtual computing resources) to perform training and / or inference using the aforementioned resources. Furthermore, one or more of the software and / or hardware resources described above may be configured to allow a user to train or perform services that infer information, such as image recognition, speech recognition, or other artificial intelligence services.

[0231] Data center 1300 may include one or more components, such as one or more central processing units (CPUs), one or more graphics processing units (GPUs), one or more isolated trusted execution environments (TEEs), one or more interconnects for multi-GPU communication, one or more data processing units (DPUs), and one or more network interface chips (NICs).

[0232] Example network environment

[0233] A network environment suitable for implementing embodiments of this disclosure may include one or more client devices, servers, network-attached storage (NAS), other backend devices, and / or other device types. Client devices, servers, and / or other device types (e.g., each device) may be... Figure 12 This is implemented on one or more instances of computing device 1200—for example, each device may include similar components, features, and / or functions of computing device 1200. Furthermore, in the case of implementing backend devices (e.g., servers, NAS, etc.), the backend devices may be included as part of data center 1300, examples of which are described herein. Figure 13 To describe in more detail.

[0234] Components of a network environment can communicate with each other via a network, which can be wired, wireless, or both. A network can include multiple networks or one of multiple networks. For example, a network can include one or more wide area networks (WANs), one or more local area networks (LANs), one or more public networks (such as the Internet and / or the Public Switched Telephone Network (PSTN)), and / or one or more private networks. Where the network includes a wireless telecommunications network, components such as base stations, communication towers, or even access points (and other components) can provide wireless connectivity.

[0235] A compatible network environment may include one or more peer-to-peer network environments (in which case the server may not be included in the network environment) and one or more client-server network environments (in which case one or more servers may be included in the network environment). In a peer-to-peer network environment, the functionality described herein for the server can be implemented on any number of client devices.

[0236] In at least one embodiment, the network environment may include one or more cloud-based network environments, distributed computing environments, combinations thereof, etc. The cloud-based network environment may include a framework layer, a job scheduler, a resource manager, and a distributed file system implemented on one or more servers, which may include one or more core network servers and / or edge servers. The framework layer may include a framework supporting software at the software layer and / or application at the application layer. The software or application may respectively include network-based service software or applications. In embodiments, one or more client devices may use the network-based service software or applications (e.g., by accessing the service software and / or applications via one or more application programming interfaces (APIs)). The framework layer may be, but is not limited to, a free and open-source software network application framework that can use a distributed file system for large-scale data processing (e.g., "big data").

[0237] A cloud-based network environment can provide cloud computing and / or cloud storage for any combination of the computing and / or data storage functions (or one or more portions thereof) described herein. Any of these different functions can be distributed across multiple locations from a central or core server (e.g., distributed across one or more data centers at the state, region, country, global, etc.). The core server may assign at least a portion of the functionality to the edge server if the connection to the user (e.g., a client device) is relatively close to the edge server. A cloud-based network environment can be private (e.g., limited to a single organization), public (e.g., available to many organizations), and / or a combination thereof (e.g., a hybrid cloud environment).

[0238] (One or more) client devices may include the information described in this article. Figure 12 At least some of the components, features, and functions of the described (one or more) example computing device 1200. By way of example and not limitation, the client device may be implemented as a personal computer (PC), laptop computer, mobile device, smartphone, tablet computer, smartwatch, wearable computer, personal digital assistant (PDA), MP3 player, virtual reality headset, global positioning system (GPS) or device, video player, camera, surveillance equipment or system, vehicle, ship, spacecraft, virtual machine, drone, robot, handheld communication device, hospital equipment, gaming equipment or system, entertainment system, vehicle computer system, embedded system controller, remote control, electrical appliance, consumer electronics device, workstation, edge device, any combination of these depicted devices, or any other suitable device.

[0239] This disclosure can be described in the general context of machine-usable instructions or computer code, including computer-executable instructions such as program modules, which are executed by a computer or other machine such as a personal digital assistant or other handheld device. Typically, a program module, including routines, programs, objects, components, data structures, etc., refers to code that performs a specific task or implements a specific abstract data type. This disclosure can be practiced in a wide variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. This disclosure can also be practiced in distributed computing environments where tasks are performed by remote processing devices linked via a communication network.

[0240] As used herein, the phrase "and / or" relating to two or more elements should be interpreted as referring to only one element or a combination of elements. For example, "element A, element B, and / or element C" can 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 element A, B, and C. Furthermore, "at least one of element A or element B" can 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. Further, "at least one of element A and element B" can 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.

[0241] The subject matter of this disclosure is described in detail herein to satisfy statutory requirements. However, the description itself is not intended to limit the scope of this disclosure. Rather, the discloser has envisioned that the claimed subject matter may be embodied in other ways to include steps different from or similar combinations of steps described herein in conjunction with other current or future techniques. Moreover, although the terms “step” and / or “box” may be used herein to imply different elements of the method employed, these terms should not be construed as implying any particular order among or between the various steps disclosed herein, unless the order of the steps is explicitly described.

[0242] Example Terms

[0243] A: A method comprising: generating one or more input markers, the one or more input markers representing one or more first points associated with a road sign depicted by an image associated with a map; generating one or more embeddings associated with the image; generating one or more output markers representing one or more second points associated with the road sign using one or more machine learning models based at least on the one or more input markers and the one or more embeddings; generating a line representation of the road sign based at least on the one or more second points; and updating the map to include labels associated with the road sign based at least on the line representation.

[0244] B: The method described in paragraph A further includes at least one of the following: receiving input data representing the one or more first points associated with the road sign; or determining the one or more first points associated with the road sign based at least on analysis of at least one of the map or the image.

[0245] C: According to the method described in paragraph A or paragraph B, generating the one or more output tags comprises: using the one or more machine learning models, based at least on the one or more input tags and the one or more embeddings, generating one or more first output tags representing a first portion of the one or more second points; and using the one or more machine learning models, based at least on the one or more first output tags, generating one or more second output tags representing a second portion of the one or more second points.

[0246] D: The method according to any one of paragraphs A and C further comprises: generating one or more image tags associated with the image using the one or more machine learning models based at least on the one or more input tags and the one or more embeddings, wherein generating the line indicates further based at least on the one or more image tags.

[0247] E: The method according to any one of paragraphs AD further includes: appending the one or more input tokens to one or more learnable tokens to generate one or more appended input tokens, wherein the generation of the one or more output tokens is based at least on the one or more appended input tokens and the one or more embeddings.

[0248] F: The method according to any one of paragraphs AE further includes: determining one or more categories associated with the one or more second points based at least on the one or more output markers, wherein generating the line represents further at least based on the one or more categories.

[0249] G: The method according to any one of paragraphs AF further comprises: using one or more decoders to generate one or more coordinates associated with the one or more second points in the image, based at least on the one or more output markers, wherein generating the line represents at least based on the one or more coordinates.

[0250] H: The method according to any one of paragraphs AG further comprises: generating a heatmap associated with the road sign based at least on at least one of the one or more output tags or one or more image tags associated with the image, wherein generating the line represents further at least based on the heatmap.

[0251] 1: A data center comprising: one or more central processing units (CPUs); one or more graphics processing units (GPUs); one or more isolated trusted execution environments (TEEs); one or more interconnects for multi-GPU communication; one or more data processing units (DPUs); one or more network interface chips (NICs); wherein one or more components of the data center are configured to: determine one or more first points associated with traffic features from sensor data representations corresponding to a map; determine one or more second points associated with the traffic features using one or more machine learning models based at least on input data associated with the one or more first points and the sensor data representations; generate a representation of the traffic features based at least on the one or more second points; and update the map to include information associated with the traffic features based at least on the representations.

[0252] J: According to the data center described in paragraph I, the one or more components are further configured to: generate one or more input tags at least based on the one or more first points, and generate one or more embeddings at least based on the sensor data representation, wherein the input data is associated with the one or more input tags and the one or more embeddings.

[0253] K: According to the data center described in paragraph I or paragraph J, wherein the one or more components are further configured to: append the one or more input tags to one or more learnable tags to generate one or more appended input tags, wherein the input data is associated with the one or more appended input tags and the one or more embeddings.

[0254] L: According to any one of paragraphs IK, the data center, wherein determining the one or more second points associated with the traffic feature comprises: generating one or more output tags using the one or more machine learning models based at least on the input data; and determining the one or more second points associated with the traffic feature based at least on the one or more output tags.

[0255] M: The data center according to any one of paragraphs IL, wherein the one or more components are further configured to perform at least one of the following: receiving one or more inputs representing the one or more first points associated with the traffic feature; or determining the one or more first points associated with the traffic feature based at least on analysis of at least one of the map or the sensor data representation.

[0256] N: According to any one of the paragraphs IM, the data center, wherein determining the one or more second points associated with the traffic feature comprises: using the one or more machine learning models, based at least on the input data, to determine at least a first portion of the one or more second points; and using the one or more machine learning models, based at least on second input data associated with the at least first portion of the one or more second points, to determine at least a second portion of the one or more second points.

[0257] O: The data center according to any one of paragraphs IN, wherein the one or more components are further configured to: determine one or more classifications associated with the one or more second points using the one or more machine learning models, at least based on the input data, wherein the representation is further generated at least based on the one or more classifications.

[0258] P: According to any one of paragraphs 10, the data center, wherein the one or more components are further configured to: determine a heatmap associated with the traffic features using the one or more machine learning models, at least based on the input data, wherein the representation is further generated at least based on the heatmap.

[0259] Q: According to any one of the paragraphs in the IP, in the data center, wherein: the traffic features include road signs represented by the sensor data representation corresponding to the map; the one or more processors are further configured to determine the sign type associated with the road signs based at least on the sensor data representation; and the map is further updated to indicate the sign type.

[0260] R: A data center according to any one of paragraphs IQ, wherein the data center is included in or used in combination with at least one of the following: a control system for autonomous or semi-autonomous machines; a perception system for autonomous or semi-autonomous machines; a system for performing one or more simulation operations; a system for performing one or more digital twin operations; a system for performing optical transmission simulation; a system for performing collaborative content creation of 3D assets; a system for providing one or more cloud gaming applications; a system for performing one or more deep learning operations; a system implemented using edge devices; a system implemented using robots; a system for performing one or more generative AI operations; a system for performing one or more large language model LLM operations. Systems for performing operations; systems for performing operations using one or more visual language models (VLMs); systems for performing operations using one or more multimodal language models; systems for performing one or more conversational AI operations; systems for generating synthetic data; systems for presenting at least one of virtual reality content, augmented reality content, or mixed reality content; systems for implementing one or more multimodal language models; systems for using or deploying one or more inference microservices; systems that include deploying one or more machine learning models and OS-level virtualization packages (e.g., containers) within services or microservices; systems that include one or more virtual machines (VMs); systems that are at least partially implemented in a data center; or systems that are at least partially implemented using cloud computing resources.

[0261] S: One or more processors, comprising: processing circuitry for generating a line representation associated with traffic features represented by a map, wherein the line representation is generated based at least on: one or more encoders of one or more machine learning models generating one or more input tags associated with one or more first points of the traffic features and one or more image embeddings associated with an image of the traffic features; and one or more decoders of the one or more machine learning models processing the one or more input tags and the one or more embeddings to determine one or more second points associated with the line representation.

[0262] T: According to paragraph S, one or more processors, wherein the one or more processors include at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing one or more analog operations; a system for performing one or more digital twin operations; a system for performing optical transmission simulation; a system for performing collaborative content creation of 3D assets; a system for providing one or more cloud gaming applications; a system for performing one or more deep learning operations; a system implemented using an edge device; a system implemented using a robot; a system for performing one or more generative AI operations; a system for performing operations using one or more large language model LLMs. A system; a system for performing operations using one or more visual language models (VLMs); a system for performing operations using one or more multimodal language models; a system for performing one or more conversational AI operations; a system for generating synthetic data; a system for presenting at least one of virtual reality content, augmented reality content, or mixed reality content; a system for implementing one or more multimodal language models; a system for using or deploying one or more inference microservices; a system that includes deploying one or more machine learning models and OS-level virtualization packages (e.g., containers) within services or microservices; a system that includes one or more virtual machines (VMs); a system that is at least partially implemented in a data center; or a system that is at least partially implemented using cloud computing resources.

Claims

1. A method comprising: Generate one or more input markers, the one or more input markers representing one or more first points associated with road signs depicted in an image associated with the map; Generate one or more embeddings associated with the image; Using one or more machine learning models and based at least on the one or more input tags and the one or more embeddings, generate one or more output tags representing one or more second points associated with the road sign; The line representation of the road sign is generated based on at least one or more of the second points; as well as The map is updated based at least on the lines to include labels associated with the road signs.

2. The method of claim 1, further comprising at least one of the following: Receive input data representing the one or more first points associated with the road sign; or Based at least on the analysis of at least one of the maps or images, the one or more first points associated with the road sign are determined.

3. The method of claim 1, wherein generating the one or more output tags comprises: Using the one or more machine learning models and based at least on the one or more input tags and the one or more embeddings, generate one or more first output tags representing the first part of the one or more second points; as well as Using the one or more machine learning models and based at least on the one or more first output tags, generate one or more second output tags representing a second part of the one or more second points.

4. The method according to claim 1, further comprising: Using the one or more machine learning models and based at least on the one or more input tags and the one or more embeddings, generate one or more image tags associated with the image. The generation of the line indicates that it is based on at least one or more image tags.

5. The method of claim 1, further comprising: The one or more input tags are appended to one or more learnable tags to generate one or more appended input tags. The generation of the one or more output tags is based at least on the one or more attached input tags and the one or more embeddings.

6. The method of claim 1, further comprising: Based at least on the one or more output tags, determine one or more categories associated with the one or more second points. The generation of the line indicates that it is further based on at least one or more classifications.

7. The method of claim 1, further comprising: Using one or more decoders and based at least on the one or more output markers, generate one or more coordinates associated with the one or more second points in the image. The line is generated based on at least one or more coordinates.

8. The method of claim 1, further comprising: A heatmap associated with the road sign is generated based on at least one of the one or more output tags or one or more image tags associated with the image. The generation of the line indicates that it is further based at least on the heatmap.

9. A data center, comprising: One or more central processing units (CPUs); One or more graphics processing units (GPUs); One or more isolated Trusted Execution Environments (TEEs); One or more interconnects used for multi-GPU communication; One or more Data Processing Units (DPUs); One or more Network Interface Chips (NICs); One or more components of the data center are used for: Identify one or more first points associated with traffic features from the sensor data representation corresponding to the map; Using one or more machine learning models and based at least on input data associated with the one or more first points and the sensor data representation, determine one or more second points associated with the traffic feature; A representation of the traffic feature is generated based on at least one or more of the second points; as well as The map is updated at least based on the representation to include information associated with the traffic features.

10. The data center of claim 9, wherein one or more components are further configured to: One or more input tags are generated based on at least one or more first points, and one or more embeddings are generated based on at least the sensor data representation. The input data is associated with one or more input tags and one or more embeddings.

11. The data center of claim 10, wherein one or more components are further configured to: The one or more input tags are appended to one or more learnable tags to generate one or more appended input tags. The input data is associated with one or more attached input tags and one or more embeddings.

12. The data center of claim 9, wherein determining the one or more second points associated with the traffic feature comprises: Using one or more machine learning models and based at least on the input data, generate one or more output labels; as well as Based on at least one or more output tags, determine one or more second points associated with the traffic feature.

13. The data center of claim 9, wherein one or more components are further configured to perform at least one of the following: Receive one or more inputs representing the one or more first points associated with the traffic feature; or The one or more first points associated with the traffic feature are determined based on at least one of the analysis of the map or the sensor data representation.

14. The data center of claim 9, wherein determining the one or more second points associated with the traffic feature comprises: Using the one or more machine learning models and based at least on the input data, determine at least a first portion of the one or more second points; as well as The at least second portion of the one or more second points is determined using the one or more machine learning models and based at least on second input data associated with the at least first portion of the one or more second points.

15. The data center of claim 9, wherein one or more components are further configured to: Using the one or more machine learning models and based at least on the input data, determine one or more classifications associated with the one or more second points. The representation is further generated based on at least one or more of the classifications.

16. The data center of claim 9, wherein one or more components are further configured to: Using one or more machine learning models and based at least on the input data, determine heatmaps associated with the traffic features. The representation is further generated based at least on the heatmap.

17. The data center according to claim 9, wherein: The traffic features include road markings represented by the sensor data corresponding to the map; The one or more processors are further configured to determine, at least based on the sensor data representation, the sign type associated with the road sign; and The map is further updated to indicate the identifier type.

18. The data center of claim 9, wherein the data center is included in at least one of the following or in combination with at least one of the following: Control systems for autonomous or semi-autonomous machines; Sensing systems for autonomous or semi-autonomous machines; A system for performing one or more simulation operations; A system for performing one or more digital twin operations; A system for performing optical transmission simulation; A system for collaborative content creation of 3D assets; A system that provides one or more cloud gaming applications; A system for performing one or more deep learning operations; Systems implemented using edge devices; Systems implemented using robots; A system for performing one or more generative artificial intelligence (AI) operations; A system for performing operations using one or more large language model LLMs; A system for performing operations using one or more visual language models (VLMs); A system for performing operations using one or more multimodal language models; A system for performing one or more conversational AI operations; A system for generating synthetic data; A system for presenting at least one of virtual reality content, augmented reality content, or mixed reality content; A system that implements one or more multimodal language models; A system that uses or deploys one or more inference microservices; A system that includes a service or microservice that deploys one or more machine learning models and an operating system (OS) level virtualization package (e.g., a container); A system containing one or more virtual machines (VMs); A system that is at least partially implemented in a data center; or A system that utilizes cloud computing resources at least in part.

19. One or more processors, comprising: Processing circuitry for generating line representations associated with traffic features represented by a map, wherein the line representations are generated based on at least the following: One or more encoders of one or more machine learning models generate one or more input tags associated with one or more first points of the traffic feature and one or more image embeddings associated with the image of the traffic feature; as well as One or more decoders of the one or more machine learning models process the one or more input tags and the one or more embeddings to determine one or more second points associated with the line representation.

20. The processor of claim 19 or more, wherein the processor is included in at least one of the following: Control systems for autonomous or semi-autonomous machines; Sensing systems for autonomous or semi-autonomous machines; A system for performing one or more simulation operations; A system for performing one or more digital twin operations; A system for performing optical transmission simulation; A system for collaborative content creation of 3D assets; A system that provides one or more cloud gaming applications; A system for performing one or more deep learning operations; Systems implemented using edge devices; Systems implemented using robots; A system for performing one or more generative AI operations; A system for performing operations using one or more large language model LLMs; A system for performing operations using one or more visual language models (VLMs); A system for performing operations using one or more multimodal language models; A system for performing one or more conversational AI operations; A system for generating synthetic data; A system for presenting at least one of virtual reality content, augmented reality content, or mixed reality content; A system that implements one or more multimodal language models; A system that uses or deploys one or more inference microservices; A system that includes a service or microservice that deploys one or more machine learning models and OS-level virtualization packages (such as containers); A system containing one or more virtual machines (VMs); A system that is at least partially implemented in a data center; or A system that utilizes cloud computing resources at least in part.

Images