Abstract: Values of pixels in an image are mapped to a binary space using a first function that preserves characteristics of values of the pixels. Labels are iteratively assigned to the pixels in the image in parallel based on a second function. The label assigned to each pixel is determined based on values of a set of nearest-neighbor pixels. The first function is trained to map values of pixels in a set of training images to the binary space and the second function is trained to assign labels to the pixels in the set of training images. Considering only the nearest neighbors in the inference scheme results in a computational complexity that is independent of the size of the solution space and produces sufficient approximations of the true distribution when the solution for each pixel is most likely found in a small subset of the set of potential solutions.

Abstract: A first and second image of a scene are captured. Each of a plurality of pixels in the first image is associated with a disparity value. An image patch associated with each of the plurality of pixels of the first image and the second image is mapped into a binary vector. Thus, values of pixels in an image are mapped to a binary space using a function that preserves characteristics of values of the pixels. The difference between the binary vector associated with each of the plurality of pixels of the first image and its corresponding binary vector in the second image designated by the disparity value associated with each of the plurality of pixels of the first image is determined. Based on the determined difference between binary vectors, correspondence between the plurality of pixels of the first image and the second image is established.

Abstract: A method includes capturing a first image and a second image of a scene using at least one imaging camera of an imaging system. The first image and the second image form a stereo image pair and each comprises a plurality of pixels. Each of the plurality of pixels in the second image is initialized with a disparity hypothesis. Matching costs of the disparity hypothesis for each of the plurality of pixels in the second image are recursively determined, from an image tile of a smaller pixel size to an image tile of a larger pixel size, to generate an initial tiled disparity map including a plurality of image tiles. After refining the disparity value estimate of each image tile and including a slant hypothesis, a final disparity estimate for each pixel of the image is generated.

Abstract: An electronic device estimates a pose of one or more subjects in an environment based on estimating a correspondence between a data volume containing a data mesh based on a current frame captured by a depth camera and a reference volume containing a plurality of fused prior data frames based on spectral embedding and performing bidirectional non-rigid matching between the reference volume and the current data frame to refine the correspondence so as to support location-based functionality. The electronic device predicts correspondences between the data volume and the reference volume based on spectral embedding. The correspondences provide constraints that accelerate the convergence between the data volume and the reference volume. By tracking changes between the current data mesh frame and the reference volume, the electronic device avoids tracking failures that can occur when relying solely on a previous data mesh frame.

Abstract: An electronic device estimates a depth map of an environment based on stereo depth images captured by depth cameras having exposure times that are offset from each other in conjunction with illuminators pulsing illumination patterns into the environment. A processor of the electronic device matches small sections of the depth images from the cameras to each other and to corresponding patches of immediately preceding depth images (e.g., a spatio-temporal image patch “cube”). The processor computes a matching cost for each spatio-temporal image patch cube by converting each spatio-temporal image patch into binary codes and defining a cost function between two stereo image patches as the difference between the binary codes. The processor minimizes the matching cost to generate a disparity map, and optimizes the disparity map by rejecting outliers using a decision tree with learned pixel offsets and refining subpixels to generate a depth map of the environment.

Abstract: A first and second image of a scene are captured. Each of a plurality of pixels in the first image is associated with a disparity value. An image patch associated with each of the plurality of pixels of the first image and the second image is mapped into a binary vector. Thus, values of pixels in an image are mapped to a binary space using a function that preserves characteristics of values of the pixels. The difference between the binary vector associated with each of the plurality of pixels of the first image and its corresponding binary vector in the second image designated by the disparity value associated with each of the plurality of pixels of the first image is determined. Based on the determined difference between binary vectors, correspondence between the plurality of pixels of the first image and the second image is established.

Abstract: Values of pixels in an image are mapped to a binary space using a first function that preserves characteristics of values of the pixels. Labels are iteratively assigned to the pixels in the image in parallel based on a second function. The label assigned to each pixel is determined based on values of a set of nearest-neighbor pixels. The first function is trained to map values of pixels in a set of training images to the binary space and the second function is trained to assign labels to the pixels in the set of training images. Considering only the nearest neighbors in the inference scheme results in a computational complexity that is independent of the size of the solution space and produces sufficient approximations of the true distribution when the solution for each pixel is most likely found in a small subset of the set of potential solutions.

Abstract: Techniques for determining depth for a visual content item using machine-learning classifiers include obtaining a visual content item of a reference light pattern projected onto an object, and determining shifts in locations of pixels relative to other pixels representing the reference light pattern. Disparity, and thus depth, for pixels may be determined by executing one or more classifiers trained to identify disparity for pixels based on the shifts in locations of the pixels relative to other pixels of a visual content item depicting in the reference light pattern. Disparity for pixels may be determined using a visual content item of a reference light pattern projected onto an object without having to match pixels between two visual content items, such as a reference light pattern and a captured visual content item.

Abstract: Correspondences in content items may be determined using a trained decision tree to detect distinctive matches between portions of content items. The techniques described include determining a first group of patches associated with a first content item and processing a first patch based at least partly on causing the first patch to move through a decision tree, and determining a second group of patches associated with a second content item and processing a second patch based at least partly on causing the second patch to move through the decision tree. The techniques described include determining that the first patch and the second patch are associated with a same leaf node of the decision tree and determining that the first patch and the second patch are corresponding patches based at least partly on determining that the first patch and the second patch are associated with the same leaf node.

Abstract: A device identifier (ID) is received that is associated with a request by a particular device to access media content hosted by a system remote from the particular device, the device ID being a device ID of the particular device. Based on the device ID, one or more capabilities of the particular device are determined relating to use of the media content. Access to the media content is based on the one or more capabilities of the particular device.

Abstract: Correspondences in content items may be determined using a trained decision tree to detect distinctive matches between portions of content items. The techniques described include determining a first group of patches associated with a first content item and processing a first patch based at least partly on causing the first patch to move through a decision tree, and determining a second group of patches associated with a second content item and processing a second patch based at least partly on causing the second patch to move through the decision tree. The techniques described include determining that the first patch and the second patch are associated with a same leaf node of the decision tree and determining that the first patch and the second patch are corresponding patches based at least partly on determining that the first patch and the second patch are associated with the same leaf node.

Abstract: Techniques for determining depth for a visual content item using machine-learning classifiers include obtaining a visual content item of a reference light pattern projected onto an object, and determining shifts in locations of pixels relative to other pixels representing the reference light pattern. Disparity, and thus depth, for pixels may be determined by executing one or more classifiers trained to identify disparity for pixels based on the shifts in locations of the pixels relative to other pixels of a visual content item depicting in the reference light pattern. Disparity for pixels may be determined using a visual content item of a reference light pattern projected onto an object without having to match pixels between two visual content items, such as a reference light pattern and a captured visual content item.

Abstract: Filtering sensor data is described, for example, where filters conditioned on a local appearance of the signal are predicted by a machine learning system, and used to filter the sensor data. In various examples the sensor data is a stream of noisy video image data and the filtering process denoises the video stream. In various examples the sensor data is a depth image and the filtering process refines the depth image which may then be used for gesture recognition or other purposes. In various examples the sensor data is one dimensional measurement data from an electric motor and the filtering process denoises the measurements. In examples the machine learning system comprises a random decision forest where trees of the forest store filters at their leaves. In examples, the random decision forest is trained using a training objective with a data dependent regularization term.

Abstract: A signal encoding an infrared (IR) image including a plurality of IR pixels is received from an IR camera. Each IR pixel specifies one or more IR parameters of that IR pixel. IR-skin pixels that image a human hand are identified in the IR image. For each IR-skin pixel, a depth of a human hand portion imaged by that IR-skin pixel is estimated based on the IR parameters of that IR-skin pixel. A skeletal hand model including a plurality of hand joints is derived. Each hand joint is defined with three independent position coordinates inferred from the estimated depths of each human hand portion.

Abstract: A signal encoding an infrared (IR) image including a plurality of IR pixels is received from an IR camera. Each IR pixel specifies one or more IR parameters of that IR pixel. IR-skin pixels that image a human hand are identified in the IR image. For each IR-skin pixel, a depth of a human hand portion imaged by that IR-skin pixel is estimated based on the IR parameters of that IR-skin pixel. A skeletal hand model including a plurality of hand joints is derived. Each hand joint is defined with three independent position coordinates inferred from the estimated depths of each human hand portion.

Abstract: A system for protecting the digital rights of content owners allows digital media to be delivered to only those media rendering client devices that have been approved for the media content. Before delivering requested media, the media service provider may determine whether the media rendering client device that requested the media is the type of device that is authorized to receive the request media. If it is, the media service provider may transmit the media to a middleman server over a network (such as the Internet). A middleman server may then serve the media to the client device over a local network. By allowing the media content to be distributed to approved devices only, the media service provider can prevent a user from using the media in a way that is not authorized by the content owner.

Abstract: A method of sensing depth using an infrared camera. In an example method, an infrared image of a scene is received from an infrared camera. The infrared image is applied to a trained machine learning component which uses the intensity of image elements to assign all or some of the image elements a depth value which represents the distance between the surface depicted by the image element and the infrared camera. In various examples, the machine line component comprises one or more random decision forests.

Abstract: Filtering sensor data is described, for example, where filters conditioned on a local appearance of the signal are predicted by a machine learning system, and used to filter the sensor data. In various examples the sensor data is a stream of noisy video image data and the filtering process denoises the video stream. In various examples the sensor data is a depth image and the filtering process refines the depth image which may then be used for gesture recognition or other purposes. In various examples the sensor data is one dimensional measurement data from an electric motor and the filtering process denoises the measurements. In examples the machine learning system comprises a random decision forest where trees of the forest store filters at their leaves. In examples, the random decision forest is trained using a training objective with a data dependent regularization term.

Abstract: A method of sensing depth using an infrared camera. In an example method, an infrared image of a scene is received from an infrared camera. The infrared image is applied to a trained machine learning component which uses the intensity of image elements to assign all or some of the image elements a depth value which represents the distance between the surface depicted by the image element and the infrared camera. In various examples, the machine line component comprises one or more random decision forests.