Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image from a video file may be enhanced using a trained neural network applying the trained neural network on the lower-resolution image to remove artifacts by removing artifacts by generating, using a layer of the trained neural network, a residual value based on the proper subset of the received image and at least one corresponding image portion of a preceding lower resolution image in the video file and at least one corresponding image portion of a subsequent lower resolution image in the video file, upscale the lower-resolution image using bilinear upsampling, and combine the upscaled received image and residual value to generate an enhanced image.

Abstract: An input video stream corresponding to input video content may be received over one or more networks. The input video stream may include a first image frame. An input visual feature of the first image frame may be matched to a reference visual feature of a reference image. The reference image may have a higher image quality than the first image frame. A replacement visual feature may be generated for the input visual feature. The replacement visual feature may be generated based at least in part on the reference visual feature. An enhanced image frame corresponding to the first image frame may be generated by at least replacing the input visual feature in the first image frame with the replacement visual feature. An enhanced video stream may be provided. The enhanced image frame may be a substitute for the first image frame in the enhanced video stream.

Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image from a video file may be enhanced using a trained neural network applying the trained neural network on the lower-resolution image to remove artifacts by removing artifacts by generating, using a layer of the trained neural network, a residual value based on the proper subset of the received image and at least one corresponding image portion of a preceding lower resolution image in the video file and at least one corresponding image portion of a subsequent lower resolution image in the video file, upscale the lower-resolution image using bilinear upsampling, and combine the upscaled received image and residual value to generate an enhanced image.

Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image from a video file may be enhanced using a trained neural network applying the trained neural network on the lower-resolution image to remove artifacts by generating, using a layer of the trained neural network, a residual value based on the proper subset of the received image and at least one corresponding image portion of a previously generated higher resolution image in the video file, upscaling the received image using bilinear upsampling, and combining the upscaled received image and residual value to generate an enhanced image.

Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image, for example from a video file, may be enhanced using a trained neural network by applying the trained neural network to enhance a middle lower-resolution image of a plurality of lower-resolution images using a generator with filters of a generative adversary network. In some examples, a plurality of sequential feature processing acts are performed on the lower-resolution images to generate a residual which is added to a filtered version of one of the lower-resolution images to generate an enhanced image.

Abstract: A data processing pipeline controller receives a request, from a data iterator associated with a machine learning model, for a data output of a module in the data processing pipeline, wherein each module in the data processing pipeline has an associated cache. The controller determines whether a data output of the module is stored in the associated cache and responsive to the data output being stored in the associated cache, provides the data output from the associated cache to the data iterator without processing data through the module.

Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image from a video file may be enhanced using a trained neural network applying the trained neural network on the lower-resolution image to remove artifacts by removing artifacts by generating, using a layer of the trained neural network, a residual value based on the proper subset of the received image and at least one corresponding image portion of a preceding lower resolution image in the video file and at least one corresponding image portion of a subsequent lower resolution image in the video file, upscale the lower-resolution image using bilinear upsampling, and combine the upscaled received image and residual value to generate an enhanced image.

Abstract: Techniques for enhancing an image are described. For example, a lower-resolution image from a video file may be enhanced using a trained neural network applying the trained neural network on the lower-resolution image to remove artifacts by generating, using a layer of the trained neural network, a residual value based on the proper subset of the received image and at least one corresponding image portion of a previously generated higher resolution image in the video file, upscaling the received image using bilinear upsampling, and combining the upscaled received image and residual value to generate an enhanced image.

Abstract: Techniques for enhancing an image are described.

Abstract: The disclosure provides an approach for detecting and localizing action in video. In one embodiment, an action detection application receives training video sequences and associated eye gaze fixation data collected from a sample of human viewers. Using the training video sequences and eye gaze data, the action detection application learns a model which includes a latent regions potential term that measures the compatibility of latent spatio-temporal regions with the model, as well as a context potential term that accounts for contextual information that is not directly produced by the appearance and motion of the actor. The action detection application may train this model in, e.g., the latent structural SVM framework by minimizing a cost function which encodes the cost of an incorrect action label prediction and a mislocalization of the eye gaze. During training and thereafter, inferences using the model may be made using an efficient dynamic programming algorithm.

Abstract: The disclosure provides an approach for detecting and localizing action in video. In one embodiment, an action detection application receives training video sequences and associated eye gaze fixation data collected from a sample of human viewers. Using the training video sequences and eye gaze data, the action detection application learns a model which includes a latent regions potential term that measures the compatibility of latent spatio-temporal regions with the model, as well as a context potential term that accounts for contextual information that is not directly produced by the appearance and motion of the actor. The action detection application may train this model in, e.g., the latent structural SVM framework by minimizing a cost function which encodes the cost of an incorrect action label prediction and a mislocalization of the eye gaze. During training and therafter, inferences using the model may be made using an efficient dynamic programming algorithm.