Abstract: Approaches are presented for training an inverse graphics network. An image synthesis network can generate training data for an inverse graphics network. In turn, the inverse graphics network can teach the synthesis network about the physical three-dimensional (3D) controls. Such an approach can provide for accurate 3D reconstruction of objects from 2D images using the trained inverse graphics network, while requiring little annotation of the provided training data. Such an approach can extract and disentangle 3D knowledge learned by generative models by utilizing differentiable renderers, enabling a disentangled generative model to function as a controllable 3D “neural renderer,” complementing traditional graphics renderers.

Abstract: A transceiver circuit includes a receiver front end utilizing a ring oscillator, and a transmitter front end utilizing a pass-gate circuit in a first feedback path across a last-stage driver circuit. The transceiver circuit provides low impedance at low frequency and high impedance at high frequency, and desirable peaking behavior.

Abstract: Estimating a three-dimensional (3D) pose of an object, such as a hand or body (human, animal, robot, etc.), from a 2D image is necessary for human-computer interaction. A hand pose can be represented by a set of points in 3D space, called keypoints. Two coordinates (x,y) represent spatial displacement and a third coordinate represents a depth of every point with respect to the camera. A monocular camera is used to capture an image of the 3D pose, but does not capture depth information. A neural network architecture is configured to generate a depth value for each keypoint in the captured image, even when portions of the pose are occluded, or the orientation of the object is ambiguous. Generation of the depth values enables estimation of the 3D pose of the object.

Abstract: A cable driving a large system such as cable driven machines, cable cars or tendons in a human or robot is typically modeled as a large number of small segments that are connected via joints. The two main difficulties with this model are satisfying the inextensibility constraint and handling the typically large mass ratio between the segments and the objects they connect. This disclosure introduces an effective approach to solving these problems. The introduced approach simulates the effect of a cable using a new type of distance constraint called ‘cable joint’ that changes both its attachment points and its rest length dynamically. The introduced approach models a cable connecting a series of objects, e.g., components of a robot, as a sequence of cable joints, reducing the complexity of the simulation from the order of the number of segments in the cable to the number of connected objects.

Abstract: Systems and methods for more accurate and robust determination of subject characteristics from an image of the subject. One or more machine learning models receive as input an image of a subject, and output both facial landmarks and associated confidence values. Confidence values represent the degrees to which portions of the subject's face corresponding to those landmarks are occluded, i.e., the amount of uncertainty in the position of each landmark location. These landmark points and their associated confidence values, and/or associated information, may then be input to another set of one or more machine learning models which may output any facial analysis quantity or quantities, such as the subject's gaze direction, head pose, drowsiness state, cognitive load, or distraction state.

Abstract: In various examples, systems and methods are disclosed that accurately identify driver and passenger in-cabin activities that may indicate a biomechanical distraction that prevents a driver from being fully engaged in driving a vehicle. In particular, image data representative of an image of an occupant of a vehicle may be applied to one or more deep neural networks (DNNs). Using the DNNs, data indicative of key point locations corresponding to the occupant may be computed, a shape and/or a volume corresponding to the occupant may be reconstructed, a position and size of the occupant may be estimated, hand gesture activities may be classified, and/or body postures or poses may be classified. These determinations may be used to determine operations or settings for the vehicle to increase not only the safety of the occupants, but also of surrounding motorists, bicyclists, and pedestrians.

Abstract: A system for managing virtual memory. The system includes a first processing unit configured to execute a first operation that references a first virtual memory address. The system also includes a first memory management unit (MMU) associated with the first processing unit and configured to generate a first page fault upon determining that a first page table that is stored in a first memory unit associated with the first processing unit does not include a mapping corresponding to the first virtual memory address. The system further includes a first copy engine associated with the first processing unit. The first copy engine is configured to read a first command queue to determine a first mapping that corresponds to the first virtual memory address and is included in a first page state directory. The first copy engine is also configured to update the first page table to include the first mapping.

Abstract: Systems and techniques for improving the performance of circuits while adapting to dynamic voltage drops caused by the execution of noisy instructions (e.g. high power consuming instructions) are provided. The performance is improved by slowing down the frequency of operation selectively for types of noisy instructions. An example technique controls a clock by detecting an instruction of a predetermined noisy type that is predicted to have a predefined noise characteristic (e.g. a high level of noise generated on the voltage rails of a circuit due to greater amount of current drawn by the instruction), and, responsive to the detecting, deceasing a frequency of the clock. The detecting occurs before execution of the instruction. The changing of the frequency in accordance with instruction type enables the circuits to be operated at high frequencies even if some of the workloads include instructions for which the frequency of operation is slowed down.

Abstract: In artificial neural networks, and other similar applications, there is typically a large amount of data involved that is considered sparse data. Due to the large size of the data involved in such applications, it is helpful to compress the data to save bandwidth resources when transmitting the data and save memory resources when storing the data. Introduced herein is a compression technique that selects elements with significant values from data and restructures them into a structured sparse format. By generating metadata that enforces the structured sparse format and organizing the data according to the metadata, the introduced technique not only reduces the size of the data but also consistently places the data in a particular format. As such, hardware can be simplified and optimized to process the data much faster and much more efficiently than the conventional compression techniques that rely on a non-structured sparsity format.

Abstract: In various examples, when a local user initiates an instance of a video conference application, the user may be provided with a user interface (UI) that displays an icon corresponding to the user as well as several other icons corresponding to participants in the instance of the video conference application. As the users converse, the local user may find that a particular participant is speaking loudly compared to the other remote users. The local user may then select an icon corresponding to the particular participant and move the icon away from the local user's icon in the UI. Based on moving the remote user's icon away from the local user's icon, the system may reduce the output volume of the audio data for the participant. Further, if the local user moves the participant icon closer to the local user's icon, the volume for the participant may be increased.

Abstract: Graphics processing unit (GPU) performance and power efficiency is improved using machine learning to tune operating parameters based on performance monitor values and application information. Performance monitor values are processed using machine learning techniques to generate model parameters, which are used by a control unit within the GPU to provide real-time updates to the operating parameters. In one embodiment, a neural network processes the performance monitor values to generate operating parameters in real-time.

Abstract: According to an aspect of an embodiment, operations may comprise determining a target position and orientation for a calibration board with respect to a camera of a vehicle, detecting a first position and orientation of the calibration board with respect to the camera of the vehicle, determining instructions for moving the calibration board from the first position and orientation to the target position and orientation, transmitting the instructions to a device, detecting a second position and orientation of the calibration board, determining whether the second position and orientation is within a threshold of matching the target position and orientation, and, in response to determining that the second position and orientation is within the threshold of matching the target position and orientation, capturing one or more calibration camera images using the camera and calibrating one or more sensors of the vehicle using the one or more calibration camera images.

Abstract: One aspect of the current disclosure provides a method for utilizing a plurality of memories associated with a plurality of devices in a computer system. The method includes: receiving an application-specific data set for executing a ray tracing application employing the devices; determining whether the data set is fully replicable in each memory; when the data set is not fully replicable in any of the memories, determining a maximum amount of the data set that is replicable in each memory while distributing a remaining amount of the data set across the memories; and identifying, based on application-specific information of the ray tracing application, a first subsection of the data set that corresponds to the maximum amount of the data set and a second subsection of the data set that corresponds to the remaining amount of the data set, wherein the first subsection is accessed more frequently than the second subsection.

Abstract: High dynamic range (HDR) support is provided for legacy application programs, such as games that are configured to display standard dynamic range (SDR) frames. HDR frames may be generated without modifying the legacy application program. The buffer creation process of the legacy application program is intercepted and modified before creation of the SDR format buffer so that the buffer is configured to use an upgraded SDR format having an increased bit depth compared with a conventional SDR buffer. Rather than tone mapping and quantizing rendered image data to the lower bit depth for storage in the conventional SDR buffer, the rendered image data is tone mapped and quantized for storage at the increased bit depth of the upgraded SDR buffer. Therefore, the luminance and greater dynamic range of the tone mapped data is better preserved compared with outputting conventional SDR frames.

Abstract: A neural network-based rendering technique increases temporal stability and image fidelity of low sample count path tracing by optimizing a distribution of samples for rendering each image in a sequence. A sample predictor neural network learns spatio-temporal sampling strategies such as placing more samples in dis-occluded regions and tracking specular highlights. Temporal feedback enables a denoiser neural network to boost the effective input sample count and increases temporal stability. The initial uniform sampling step typically present in adaptive sampling algorithms is not needed. The sample predictor and denoiser operate at interactive rates to achieve significantly improved image quality and temporal stability compared with conventional adaptive sampling techniques.

Abstract: When a signal glitches, logic receiving the signal may change in response, thereby charging and/or discharging nodes within the logic and dissipating power. Providing a glitch-free signal may reduce the number of times the nodes are charged and/or discharged, thereby reducing the power dissipation. A technique for eliminating glitches in a signal is to insert a storage element that samples the signal after it is done changing to produce a glitch-free output signal. The storage element is enabled by a “ready” signal having a delay that matches the delay of circuitry generating the signal. The technique prevents the output signal from changing until the final value of the signal is achieved. The output signal changes only once, typically reducing the number of times nodes in the logic receiving the signal are charged and/or discharged so that power dissipation is also reduced.

Abstract: A system and method for an on-demand shuttle, bus, or taxi service able to operate on private and public roads provides situational awareness and confidence displays. The shuttle may include ISO 26262 Level 4 or Level 5 functionality and can vary the route dynamically on-demand, and/or follow a predefined route or virtual rail. The shuttle is able to stop at any predetermined station along the route. The system allows passengers to request rides and interact with the system via a variety of interfaces, including without limitation a mobile device, desktop computer, or kiosks. Each shuttle preferably includes an in-vehicle controller, which preferably is an AI Supercomputer designed and optimized for autonomous vehicle functionality, with computer vision, deep learning, and real time ray tracing accelerators. An AI Dispatcher performs AI simulations to optimize system performance according to operator-specified system parameters.

Abstract: Often there are errors when reading data from computer memory. To detect and correct these errors, there are multiple types of error correction codes. Disclosed is an error correction architecture that creates a codeword having a data portion and an error correction code portion. Swizzling rearranges the order of bits and distributes the bits among different codewords. Because the data is redistributed, a potential memory error of up to N contiguous bits, where N for example equals 2 times the number of codewords swizzled together, only affects up to, at most, two bits per swizzled codeword. This keeps the error within the error detecting capabilities of the error correction architecture. Furthermore, this can allow improved error correction and detection without requiring a change to error correcting code generators and checkers.

Abstract: A clock data recovery circuit detects illegal decisions for received data, accumulates a phase gradient for the data, determines a number of the illegal decisions in a configured window for receiving the data, and if the number of the illegal decisions exceeds a pre-defend number in the window, applies a sum of the accumulated phase gradient and a phase increment having a sign of the accumulated phase gradient to a clock circuit for the data receiver.

Abstract: In various examples, a two-dimensional (2D) and three-dimensional (3D) deep neural network (DNN) is implemented to fuse 2D and 3D object detection results for classifying objects. For example, regions of interest (ROIs) and/or bounding shapes corresponding thereto may be determined using one or more region proposal networks (RPNs)—such as an image-based RPN and/or a depth-based RPN. Each ROI may be extended into a frustum in 3D world-space, and a point cloud may be filtered to include only points from within the frustum. The remaining points may be voxelated to generate a volume in 3D world space, and the volume may be applied to a 3D DNN to generate one or more vectors. The one or more vectors, in addition to one or more additional vectors generated using a 2D DNN processing image data, may be applied to a classifier network to generate a classification for an object.