Microsoft Technology Licensing, LLC

Richard E Moore

Mark James Finocchio

Alexandru Octavian Balan

Jamie Shotton

Andrew William Fitzgibbon

Kyungsuk David Lee

Avi Bar-Zeev

Rod G Fleck

Kathryn Stone Perez

Alex Aben-Athar Kipman

Ryan M Geiss

Andrew John Fuller

Rudy Jacobus Poot

Duncan Paul Robertson

Landon M Dyer

Mihai-Dan Budiu

Mat Cook

Jamie D J Shotton

Mark Finnochio

Attribute state classification techniques are described. In one or more implementations, one or more pixels of an image are classified by a computing device as having one or several states for one or more attributes that do not identify corresponding body parts of a user. A gesture is recognized by the computing device that is operable to initiate one or more operations of the computing device based at least in part of the state classifications of the one or more pixels of one or more attributes.

Attribute state classification

Embodiments related to detecting object information from image data collected by an image sensor are disclosed. In one example embodiment, the object information is detected by receiving a frame of image data from the image sensor and detecting a change in a threshold condition related to an object within the frame. The embodiment further comprises adjusting a setting that changes a power consumption of the image sensor in response to detecting the threshold condition.

Controlling power consumption in object tracking pipeline

Embodiments are disclosed herein that relate to generating a decision tree through graphical processing unit (GPU) based machine learning. For example, one embodiment provides a method including, for each level of the decision tree: performing, at each GPU of the parallel processing pipeline, a feature test for a feature in a feature set on every example in an example set. The method further includes accumulating results of the feature tests in local memory blocks. The method further includes writing the accumulated results from each local memory block to global memory to generate a histogram of features for every node in the level, and for each node in the level, assigning a feature having a lowest entropy in accordance with the histograms to the node.

Parallel processing machine learning decision tree training

A system and method for analyzing, summarizing, and transmitting life experiences captured using a life recorder is described. A life recorder is a recording device that continuously captures life experiences, including unanticipated life experiences, in video and/or audio recordings. In some embodiments, the video and/or audio recordings generated by a life recorder are automatically summarized, indexed, and stored for future use. By indexing and storing life recordings, a life recorder may search for and acquire life recordings generated by itself or another life recorder, thereby allowing life experiences to be shared minutes or even years later. In some embodiments, recordings generated by a life recorder may be analyzed in real-time and automatically pushed to one or more target devices. The ability to automatically and instantaneously push life recordings as live feeds to one or more target devices allows friends and family to experience one's life experience in real-time.

Life streaming

Technology is described for determining and using invariant features for computer vision. A local orientation may be determined for each depth pixel in a subset of the depth pixels in a depth map. The local orientation may an in-plane orientation, an out-out-plane orientation or both. A local coordinate system is determined for each of the depth pixels in the subset based on the local orientation of the corresponding depth pixel. A feature region is defined relative to the local coordinate system for each of the depth pixels in the subset. The feature region for each of the depth pixels in the subset is transformed from the local coordinate system to an image coordinate system of the depth map. The transformed feature regions are used to process the depth map.

Invariant features for computer vision

A computerized decision tree training system may include a distributed control processing unit configured to receive input of training data for training a decision tree. The system may further include a plurality of data batch processing units, each data batch processing unit being configured to evaluate each of a plurality of split functions of a decision tree for respective data batch of the training data, to thereby compute a partial histogram for each split function, for each datum in the data batch. The system may further include a plurality of node batch processing units configured to aggregate the associated partial histograms for each split function to form an aggregated histogram for each split function for each of a subset of frontier tree nodes and to determine a selected split function for each frontier tree node by computing the split function that produces highest information gain for the frontier tree node.

Distributed decision tree training

Synthesized body images are generated for a machine learning algorithm of a body joint tracking system. Frames from motion capture sequences are retargeted to several different body types, to leverage the motion capture sequences. To avoid providing redundant or similar frames to the machine learning algorithm, and to provide a compact yet highly variegated set of images, dissimilar frames can be identified using a similarity metric. The similarity metric is used to locate frames which are sufficiently distinct, according to a threshold distance. For realism, noise is added to the depth images based on noise sources which a real world depth camera would often experience. Other random variations can be introduced as well. For example, a degree of randomness can be added to retargeting. For each frame, the depth image and a corresponding classification image, with labeled body parts, are provided. 3-D scene elements can also be provided.

Proxy training data for human body tracking

Damage to removable or non-removable media or to content on the media may prevent access to content on the media. A processor may recognize a defect in the media, for example, due to read errors. The processor may request a 'clean' (i.e., undamaged) or replaceable version of the non-readable media content. The processor may request the content over a network. The processor may store the requested content in local storage, such as in the cache or more permanent memory. The processor may read the requested portion directly over the network or from the stored location, rather than reading the damaged portion from the removable media. Once stored locally, the clean version may be available for subsequent access. Subsequently, the processor may request previously stored content from local storage without needing to request content over the network.

Growing community of inventors

Redmond, WA, United States of America

Richard E Moore