Techniques to improve performance of matrix multiply operations are described in which a compute kernel can specify one or more element-wise operations to perform on output of the compute kernel before the output is transferred to higher levels of a processor memory hierarchy.

System, methods, and other embodiments described herein relate to semi-supervised training of a depth model for monocular depth estimation. In one embodiment, a method includes training the depth model according to a first stage that is self-supervised and that includes using first training data that comprises pairs of training images. Respective ones of the pairs including separate frames depicting a scene of a monocular video. The method includes training the depth model according to a second stage that is weakly supervised and that includes using second training data to produce depth maps according to the depth model. The second training data comprising individual images with corresponding sparse depth data. The second training data providing for updating the depth model according to second stage loss values that are based, at least in part, on the depth maps and the depth data.

System, methods, and other embodiments described herein relate to training a depth model for monocular depth estimation. In one embodiment, a method includes generating, as part of training the depth model according to a supervised training stage, a depth map from a first image of a pair of training images using the depth model. The pair of training images are separate frames depicting a scene from a monocular video. The method includes generating a transformation from the first image and a second image of the pair using a pose model. The method includes computing a supervised loss based, at least in part, on reprojecting the depth map and training depth data onto an image space of the second image according to at least the transformation. The method includes updating the depth model and the pose model according to at least the supervised loss.

An image processing method including: capturing changes in a monitored scene; and performing a sparse feature calculation on the changes in the monitored scene to obtain a sparse feature map.

Techniques to improve performance of matrix multiply operations are described in which a compute kernel can specify one or more element-wise operations to perform on output of the compute kernel before the output is transferred to higher levels of a processor memory hierarchy.

Techniques to improve performance of matrix multiply operations are described in which a compute kernel can specify one or more element-wise operations to perform on output of the compute kernel before the output is transferred to higher levels of a processor memory hierarchy.

A computer-implemented method and corresponding system for processing sensor data associated with a vehicle is provided. The sensor data may be compressed or encoded with a dictionary according to sparse approximation theory, resulting in a sparse representation of the sensor data. Processing may further comprise detecting an event associated with the vehicle, wherein an event may be an accident recorded by sensors of the vehicle providing the sensor data. The detection of the event may be based on processing of the sparse representation of the sensor data alone without decoding the sparse representation. The detection of the event may further employ machine learning methods trained to the detection of an event from the sparse representation of the sensor data, or a combination of sparse representations of sensor data originating from a plurality of vehicles or sensors.

An apparatus to facilitate processing of a sparse matrix is disclosed. The apparatus includes a plurality of processing units each comprising one or more processing elements, including logic to read operands, a multiplication unit to multiply two or more operands and a scheduler to identify operands having a zero value and prevent scheduling of the operands having the zero value at the multiplication unit.

An apparatus to facilitate processing of a sparse matrix is disclosed. The apparatus includes a plurality of processing units each comprising one or more processing elements, including logic to read operands, a multiplication unit to multiply two or more operands and a scheduler to identify operands having a zero value and prevent scheduling of the operands having the zero value at the multiplication unit.

An apparatus and method of denoising a dynamic image is provided. The dynamic image can represent a time-series of snapshot images. The dynamic image is transformed, using a sparsifying transformation, into an aggregate image and a series of transform-domain images. The transform-domain images represent kinetic information of the dynamic images (i.e., differences between the snapshots), and the aggregate image represents static information (i.e., features and structure common among the snapshots). The transform-domain images, which can be approximated using a sparse approximation method, are denoised. The denoised transform-domain images are recombined with the aggregate image using an inverse sparsifying transformation to generate a denoised dynamic image. The transform-domain images can be denoised using at least one of a principal component analysis method and a K-SVD method.