Abstract: In various examples, artificial intelligence (AI) agents can be generated to synthesize more natural motion by simulated actors in various visualizations (such as video games or simulations). AI agents may employ one or more machine learning models and techniques, such as reinforcement learning, to enable synthesis of motion with enhanced realism. The AI agent can be trained based on widely-available broadcast video data, without the need for more costly and limited motion capture data. To account for the lower quality of such video data, various techniques can be employed, such as taking into account the motion of joints, and applying physics-based constraints on the actors, resulting in higher quality, more lifelike motion.

Abstract: In various examples, to support training a deep neural network (DNN) to predict a dense representation of a 3D surface structure of interest, a training dataset is generated using a parametric mathematical modeling. A variety of synthetic 3D road surfaces may be generated by modeling a 3D road surface using varied parameters to simulate changes in road direction and lateral surface slope. In an example embodiment, a synthetic 3D road surface may be created by modeling a longitudinal 3D curve and expanding the longitudinal 3D curve to a 3D surface, and the resulting synthetic 3D surface may be sampled to form a synthetic ground truth projection image (e.g., a 2D height map). To generate corresponding input training data, a known pattern that represents which pixels may remain unobserved during 3D structure estimation may be generated and applied to a ground truth projection image to simulate a corresponding sparse projection image.

Abstract: Various embodiments and implementations of graph-neural-network (GNN)-based decoding applications are disclosed. The GNN-based decoding schemes are broadly applicable to different coding schemes, and capable of operating on both binary and non-binary codewords, in different implementations. Advantageously, the inventive GNN-based decoding is scalable, even with arbitrary block lengths, and not subject to typical limits with respect to dimensionality. Decoding performance of the inventive GNN-based techniques demonstrably matches or outpaces BCH and LDPC (both regular and 5G NR) decoding algorithms, while exhibiting improvements with respect to number of iterations required and scalability of the GNN-based approach. These inventive concepts are implemented, according to various embodiments, as methods, systems, and computer program products.

Abstract: In various examples, a machine learning model—such as a deep neural network (DNN)—may be trained to use image data and/or other sensor data as inputs to generate two-dimensional or three-dimensional trajectory points in world space, a vehicle orientation, and/or a vehicle state. For example, sensor data that represents orientation, steering information, and/or speed of a vehicle may be collected and used to automatically generate a trajectory for use as ground truth data for training the DNN. Once deployed, the trajectory points, the vehicle orientation, and/or the vehicle state may be used by a control component (e.g., a vehicle controller) for controlling the vehicle through a physical environment. For example, the control component may use these outputs of the DNN to determine a control profile (e.g., steering, decelerating, and/or accelerating) specific to the vehicle for controlling the vehicle through the physical environment.

Abstract: A technique is described for extracting or constructing a three-dimensional (3D) model from multiple two-dimensional (2D) images. In an embodiment, a foreground segmentation mask or depth field may be provided as an additional supervision input with each 2D image. In an embodiment, the foreground segmentation mask or depth field is automatically generated for each 2D image. The constructed 3D model comprises a triangular mesh topology, materials, and environment lighting. The constructed 3D model is represented in a format that can be directly edited and/or rendered by conventional application programs, such as digital content creation (DCC) tools. For example, the constructed 3D model may be represented as a triangular surface mesh (with arbitrary topology), a set of 2D textures representing spatially-varying material parameters, and an environment map. Furthermore, the constructed 3D model may be included in 3D scenes and interacts realistically with other objects.

Abstract: Apparatuses, systems, and techniques for supporting fairness of multiple context sharing cryptographic hardware. An accelerator circuit includes a copy engine (CE) with AES-GCM hardware configured to perform both encryption and authentication of data transfers for multiple applications or multiple data streams in a single application or belonging to a single user. The CE splits a data transfer of a specified size into a set of partial transfers. The CE sequentially executes the set of partial transfers using a context for a period of time (e.g., a timeslice) for an application. The CE stores in a secure memory for the application one or more data for encryption or decryption (e.g., a hash key, a block counter, etc.) computed from a last partial transfer. The one or more data for encryption or decryption are retrieved and used when data transfers for the application is resumed by the CE.

Abstract: In various examples, a current claimed set of points representative of a volume in an environment occupied by a vehicle at a time may be determined. A vehicle-occupied trajectory and at least one object-occupied trajectory may be generated at the time. An intersection between the vehicle-occupied trajectory and an object-occupied trajectory may be determined based at least in part on comparing the vehicle-occupied trajectory to the object-occupied trajectory. Based on the intersection, the vehicle may then execute the first safety procedure or an alternative procedure that, when implemented by the vehicle when the object implements the second safety procedure, is determined to have a lesser likelihood of incurring a collision between the vehicle and the object than the first safety procedure.

Abstract: Systems and methods for an efficient and robust multiprocessor-coprocessor interface that may be used between a streaming multiprocessor and an acceleration coprocessor in a GPU are provided. According to an example implementation, in order to perform an acceleration of a particular operation using the coprocessor, the multiprocessor: issues a series of write instructions to write input data for the operation into coprocessor-accessible storage locations, issues an operation instruction to cause the coprocessor to execute the particular operation; and then issues a series of read instructions to read result data of the operation from coprocessor-accessible storage locations to multiprocessor-accessible storage locations.

Abstract: In various examples, a sensor model may be learned to predict virtual sensor data for a given scene configuration. For example, a sensor model may include a deep neural network that supports generative learning—such as a generative adversarial network (GAN). The sensor model may accept an encoded representation of a scene configuration as an input using any number of data structures and/or channels (e.g., concatenated vectors, matrices, tensors, images, etc.), and may output virtual sensor data. Real-world data and/or virtual data may be collected and used to derive training data, which may be used to train the sensor model to predict virtual sensor data for a given scene configuration. As such, one or more sensor models may be used as virtual sensors in any of a variety of applications, such as in a simulated environment to test features and/or functionality of one or more autonomous or semi-autonomous driving software stacks.

Abstract: Systems and methods are disclosed for throttling memory bandwidth accessed by virtual machines (VMs). A technique for dynamically throttling the virtual computer processing units (vCPUs) assigned to a VM (tenant) controls the memory access rate of the VM. When the memory is shared by multiple VMs in a cloud-computing environment, one VM increasing its memory access rate may cause another VM to suffer memory access starvation. This behavior violates the principle of VM isolation in cloud computing. In contrast to conventional systems, a software solution for dynamically throttling the vCPUs may be implemented within a hypervisor and is therefore portable across CPU families and doesn't require specialized server-class CPU capabilities or limit the system configuration.

Abstract: Machine learning is a process that learns a model from a given dataset, where the model can then be used to make a prediction about new data. In order to reduce the costs associated with collecting and labeling real world datasets for use in training the model, computer processes can synthetically generate datasets which simulate real world data. The present disclosure improves the effectiveness of such synthetic datasets for training machine learning models used in real world applications, in particular by generating a synthetic dataset that is specifically targeted to a specified downstream task (e.g. a particular computer vision task, a particular natural language processing task, etc.).

Abstract: A framework for offline learning from a set of diverse and suboptimal demonstrations operates by selectively imitating local sequences from the dataset. At least one embodiment recovers performant policies from large manipulation datasets by decomposing the problem into a goal-conditioned imitation and a high-level goal selection mechanism.

Abstract: In various examples, a deep neural network(s) (e.g., a convolutional neural network) may be trained to detect moving and stationary obstacles from RADAR data of a three dimensional (3D) space. In some embodiments, ground truth training data for the neural network(s) may be generated from LIDAR data. More specifically, a scene may be observed with RADAR and LIDAR sensors to collect RADAR data and LIDAR data for a particular time slice. The RADAR data may be used for input training data, and the LIDAR data associated with the same or closest time slice as the RADAR data may be annotated with ground truth labels identifying objects to be detected. The LIDAR labels may be propagated to the RADAR data, and LIDAR labels containing less than some threshold number of RADAR detections may be omitted. The (remaining) LIDAR labels may be used to generate ground truth data.

Abstract: A neural network structure is separated into an odd neural network including only the odd layers and an even neural network including only the even layers. In order to allow for parallel execution, for forward propagation a second input is generated from the original input, while for backward propagation a second error gradient is generated. Parallel execution may accelerate the forward and backward propagation operations without significant change in accuracy of the model. Additionally, restructuring a single neural network into two or more parallel neural networks may reduce the total time needed for training.

Abstract: Systems and methods related to transferring (e.g., large) files over a network are disclosed. In at least one embodiment, a client-server framework establishes a QUIC connection between a server application and a client application. Source files are processed by the server application to divide the source files into a number of chunks. Differential file transfer can be implemented between the client application and the server application by comparing metadata for chunks of the source file with metadata of local chunks of a destination file already stored in a local storage associated with the client application. Missing chunks can be requested from the server application and transferred to the client application using HTTP/3 messages.

Abstract: Methods and apparatus are described for detecting anomalies in a clock signal. Example methods include sensing a clock signal that exhibits alternating phases during normal operation; responsive to sensing the start of a first phase, generating a pulse; and if the pulse terminates before sensing the end of the first phase, asserting a clock stopped detection signal. Example clock anomaly detection apparatus includes a clock signal input for coupling to a clock signal that, during normal operation, oscillates between first and second clock states. An anomaly detection output is asserted if the clock signal remains in the first clock state longer than a first phase expected duration or remains in the second clock state longer than a second phase expected duration.

Abstract: Disclosed approaches provide for interactions of secondary rays of light transport paths in a virtual environment to share lighting contributions when determining lighting conditions for a light transport path. Interactions may be shared based on similarities in characteristics (e.g., hit locations), which may define a region in which interactions may share lighting condition data. The region may correspond to a texel of a texture map and lighting contribution data for interactions may be accumulated to the texel spatially and/or temporally, then used to compute composite lighting contribution data that estimates radiance at an interaction. Approaches are also provided for reprojecting lighting contributions of interactions to pixels to share lighting contribution data from secondary bounces of light transport paths while avoiding potential over blurring.

Abstract: A multi-level contrastive training strategy for training a neural network relies on image pairs (no other labels) to learn semantic correspondences at the image level and region or pixel level. The neural network is trained using contrasting image pairs including different objects and corresponding image pairs including different views of the same object. Conceptually, contrastive training pulls corresponding image pairs closer and pushes contrasting image pairs apart. An image-level contrastive loss is computed from the outputs (predictions) of the neural network and used to update parameters (weights) of the neural network via backpropagation. The neural network is also trained via pixel-level contrastive learning using only image pairs. Pixel-level contrastive learning receives an image pair, where each image includes an object in a particular category.

Abstract: A geometric approach may be used to detect objects on a road surface. A set of points within a region of interest between a first frame and a second frame are captured and tracked to determine a difference in location between the set of points in two frames. The first frame may be aligned with the second frame and the first pixel values of the first frame may be compared with the second pixel values of the second frame to generate a disparity image including third pixels. One or more subsets of the third pixels that have a value above a first threshold may be combined, and the third pixels may be scored and associated with disparity values for each pixel of the one or more subsets of the third pixels. A bounding shape may be generated based on the scoring.

Abstract: A glitch detection device includes an oscillator to generate multiple local clocks of multiple different phases and a sampling circuit to oversample, using the multiple local clocks, a system clock to generate multiple samples of the system clock. The device further includes digital logic that in turn includes a glitch detector to monitor a variation in pulse width of the system clock based on counting the multiple samples and to report a glitch in response to detecting a variation in the pulse width that exceeds a threshold value. The digital logic further includes a loop filter coupled between the glitch detector and the oscillator. The loop filter variably adjusts the oscillator based on a frequency of each of the multiple samples to control an output frequency of each of the multiple different phases of the oscillator.