Abstract: Various types of image analysis benefit from a multi-stream architecture that allows the analysis to consider shape data. A shape stream can process image data in parallel with a primary stream, where data from layers of a network in the primary stream is provided as input to a network of the shape stream. The shape data can be fused with the primary analysis data to produce more accurate output, such as to produce accurate boundary information when the shape data is used with semantic segmentation data produced by the primary stream. A gate structure can be used to connect the intermediate layers of the primary and shape streams, using higher level activations to gate lower level activations in the shape stream. Such a gate structure can help focus the shape stream on the relevant information and reduces any additional weight of the shape stream.

Abstract: One embodiment of a method for computing a texture color includes tracing a ray cone through a graphics scene, determining at least one axis of an ellipse formed by the ray cone intersecting a plane associated with geometry within the graphics scene at a hit point, computing one or more gradients along the at least one axis of the ellipse, and computing a texture color based on the one or more gradients and a texture.

Abstract: A game-agnostic event detector can be used to automatically identify game events. Event data for detected events can be written to an event log in a form that is both human- and process-readable. Descriptive text for the event data can come from a common event dictionary that is hierarchical in nature, such that events of the same type can be correlated across different games even though the precise nature or appearance of those events may be different. The event data can be used for various purposes, such as to generate highlight videos or provide player performance feedback.

Abstract: According to an aspect of an embodiment, operations may comprise capturing, at a vehicle as the vehicle travels, LIDAR scans and camera images. The operations may further comprise selecting, at the vehicle as the vehicle travels, a subset of the LIDAR scans and the camera images that are determined to be useful for calibration. The operations may further comprise computing, at the vehicle as the vehicle travels, LIDAR-to-camera transformations for the subset of the LIDAR scans and the camera images using an optimization algorithm. The operations may further comprise calibrating, at the vehicle as the vehicle travels, one or more sensors of the vehicle based on the LIDAR-to-camera transformations.

Abstract: In various examples, a deep learning solution for path detection is implemented to generate a more abstract definition of a drivable path without reliance on explicit lane-markings—by using a detection-based approach. Using approaches of the present disclosure, the identification of drivable paths may be possible in environments where conventional approaches are unreliable, or fail—such as where lane markings do not exist or are occluded. The deep learning solution may generate outputs that represent geometries for one or more drivable paths in an environment and confidence values corresponding to path types or classes that the geometries correspond. These outputs may be directly useable by an autonomous vehicle—such as an autonomous driving software stack—with minimal post-processing.

Abstract: In various examples, sensor data representative of an image of a field of view of a vehicle sensor may be received and the sensor data may be applied to a machine learning model. The machine learning model may compute a segmentation mask representative of portions of the image corresponding to lane markings of the driving surface of the vehicle. Analysis of the segmentation mask may be performed to determine lane marking types, and lane boundaries may be generated by performing curve fitting on the lane markings corresponding to each of the lane marking types. The data representative of the lane boundaries may then be sent to a component of the vehicle for use in navigating the vehicle through the driving surface.

Abstract: A vehicle, for example, an autonomous vehicle receives signals from a global navigation satellite system (GNSS) and determines accurate location of the vehicle using the GNSS signal. The vehicle performs localization to determine the location of the vehicle as it drives. The autonomous vehicle uses sensor data and a high definition map to determine an accurate location of the autonomous vehicle. The autonomous vehicle uses accurate location of the vehicle to determine RTK corrections that is used for improving GNSS location estimates at a future location. The RTK corrections may be transmitted to other vehicles.

Abstract: During functional/normal operation of an integrated circuit including multiple independent processing elements, a selected independent processing element is taken offline and the functionality of the selected independent processing element is then tested while the remaining independent processing elements continue functional operation. To minimize voltage drops resulting from current fluctuations produced by the testing of the processing element, clocks used to synchronize operations within each partition of a processing element are staggered. This varies the toggle rate within each partition of the processing element during the testing of the processing core, thereby reducing the resulting voltage drop. This may also improve test quality within an automated test equipment (ATE) environment.

Abstract: In an embodiment, a system provides object tracking and 6D pose estimations to a robot that performs different tasks such as manipulation and navigation. In an embodiment the 6D object pose is determined using a Rao-Blackwellized particle filtering framework, where the 3-D rotation and the 3-D translation of the object is decoupled. In an embodiment, the system provides the 3-D translation of an object along with a full distribution over the 3-D rotation. In an embodiment, the 3-D rotation is determined by discretizing the rotation space, and training an autoencoder network to construct a codebook of feature embeddings for the discretized rotations. In an embodiment, the system is able to track objects with arbitrary symmetries while also maintaining adequate posterior distributions.

Abstract: Unavoidable physical phenomena, such as an alpha particle strikes, can cause soft errors in integrated circuits. Materials that emit alpha particles are ubiquitous, and higher energy cosmic particles penetrate the atmosphere and also cause soft errors. Some soft errors have no consequence, but others can cause an integrated circuit to malfunction. In some applications (e.g. driverless cars), proper operation of integrated circuits is critical to human life and safety. To minimize or eliminate the likelihood of a soft error becoming a serious malfunction, detailed assessment of individual potential soft errors and subsequent processor behavior is necessary. Embodiments of the present disclosure facilitate emulating a plurality of different, specific soft errors. Resilience may be assessed over the plurality of soft errors and application code may be advantageously engineered to improve resilience.

Abstract: A parallel processing unit (PPU) can be divided into partitions. Each partition is configured to operate similarly to how the entire PPU operates. A given partition includes a subset of the computational and memory resources associated with the entire PPU. Software that executes on a CPU partitions the PPU for an admin user. A guest user is assigned to a partition and can perform processing tasks within that partition in isolation from any other guest users assigned to any other partitions. Because the PPU can be divided into isolated partitions, multiple CPU processes can efficiently utilize PPU resources.

Abstract: Attributes of graphics objects are processed in a plurality of graphics processing pipelines. A streaming multiprocessor (SM) retrieves a first set of parameters associated with a set of graphics objects from a first set of buffers. The SM performs a first set of operations on the first set of parameters according to a first phase of processing to produce a second set of parameters stored in a second set of buffers. The SM performs a second set of operations on the second set of parameters according to a second phase of processing to produce a third set of parameters stored in a third set of buffers. One advantage of the disclosed techniques is that work is redistributed from a first phase to a second phase of graphics processing without having to copy the attributes to and retrieve the attributes from the cache or system memory, resulting in reduced power consumption.

Abstract: A bounding volume is used to approximate the space an object occupies. If a more precise understanding beyond an approximation is required, the object itself is then inspected to determine what space it occupies. Often, a simple volume (such as an axis-aligned box) is used as bounding volume to approximate the space occupied by an object. But objects can be arbitrary, complicated shapes. So a simple volume often does not fit the object very well. That causes a lot of space that is not occupied by the object to be included in the approximation of the space being occupied by the object. Hardware-based techniques are disclosed herein, for example, for efficiently using multiple bounding volumes (such as axis-aligned bounding boxes) to represent, in effect, an arbitrarily shaped bounding volume to better fit the object, and for using such arbitrary bounding volumes to improve performance in applications such as ray tracing.

Abstract: This disclosure presents a method and computer program product to denoise a ray traced scene. An apparatus for processing a ray traced scene is also disclosed. In one example, the method includes: (1) generating filtered scene data by filtering modified scene data from original scene data utilizing a spatial filter, and (2) providing a denoised ray traced scene by adjusting the filtered scene data utilizing a temporal filter. The modified and adjusted scene data can be sent to a rendering processor or system to complete rendering to generate a final scene.

Abstract: The disclosure provides features or schemes that improve a user's experience with an interactive computer product by reducing latency through late latching and late warping. The late warping can be applied by imaging hardware based on late latch inputs and is applicable for both local and cloud computing environments. In one aspect, the disclosure provides a method of operating an imaging system employing late latching and late warping. In one example the method of operating an imaging system includes: (1) rendering a rendered image based on a user input from an input device and scene data from an application engine, (2) obtaining a late latch input from the input device, (3) rendering, employing imaging hardware, a warped image by late warping at least a portion of the rendered image based on the late latch input, and (4) updating state information in the application engine with late latch and warp information.

Abstract: Devices, systems, and techniques to incorporate lighting effects into computer-generated graphics. In at least one embodiment, a virtual scene comprising a plurality of lights is rendered by randomly sampling a set of lights from among the plurality of lights prior to rendering a frame of graphics. A subset of the set of lights is selected and used to render pixels within one or more portions of the frame.

Abstract: A patch scanning display apparatus and a technique for reconstructing a target image frame on a projection surface is disclosed. The patch scanning display apparatus includes a backlight and a spatial light modulator (SLM). An optical scanning device scans the image projected by the SLM across the projection surface in accordance with a scan trajectory. A decomposition model is used to generate a set of image patches based on the target image frame and the scan trajectory. In an embodiment, the decomposition model is a projective non-negative matrix factorization model. The set of image patches are utilized to generate a modulation signal for the SLM and a binary backlight signal is then generated for each time step of the scan trajectory within a frame period to activate or deactivate the light-emitting elements of the backlight during the frame period at a high refresh rate while the projected image is scanned.

Abstract: A feed forward equalizer including a first set of filter taps having a first set of filter tap coefficients to be adapted and a second set of one or more filter taps having one or more filter tap coefficients to be constrained. The feed forward equalizer includes an adaptation component to determine a set of adapted filter tap coefficient values corresponding to the first set of filter tap coefficients and a constraint function component to determine a constrained filter tap coefficient value for the second set of the one or more filter taps having the one or more filter tap coefficients to be constrained using a constraint function based on at least a portion of the set of adapted filter tap coefficient values. The feed forward equalizer generates, based at least in part on the constrained filter tap coefficient value, an equalized signal including a set of estimated symbol values.

Abstract: In various examples, frames of a video may include a first visual object that may appear relative to a second visual object within a region of the frames. Once a relationship between the first visual object and the region is known, one or more operations may be performed on the relative region. For example, optical character recognition may be performed on the relative region where the relative region is known to contain textual information. As a result, the identification of the first visual object may serve as an anchor for determining the location of the relative region including the second visual object—thereby increasing accuracy and efficiency of the system while reducing run-time.

Abstract: Systems and methods enable optimization of a 3D model representation comprising the shape and appearance of a particular 3D scene or object. The opaque 3D mesh (e.g., vertex positions and corresponding topology) and spatially varying material attributes are jointly optimized based on image space losses to match multiple image observations (e.g., reference images of the reference 3D scene or object). A geometric topology defines faces and/or cells in the opaque 3D mesh that are visible and may be randomly initialized and optimized through training based on the image space losses. Applying the geometry topology to an opaque 3D mesh for learning the shape improves accuracy of silhouette edges and performance compared with using transparent mesh representations. In contrast with approaches that require an initial guess for the topology and/or an exhaustive testing of possible geometric topologies, the 3D model representation is learned based on image space differences without requiring an initial guess.