Microsoft Technology Licensing, LLC

A9.com, Inc.

Videosurf, Inc.

Amazon Technologies, Inc.

Eran Borenstein

Eitan Sharon

Achiezer Brandt

Mehmet Nejat Tek

Asael Moshe

Lior Delgo

Arnab Sanat Kumar Dhua

Noam Sorek

Simon John Baker

Anish Ramdas Nair

Chetan Pitambar Bhole

Erick Cantu-Paz

Zur Nehushtan

Arunkumar Devadoss

Sunaad Nataraju

Himanshu Arora

Ianir Ideses

Eduard Oks

Dotan Kaufman

Achi Brandt

Matan Goldman

Imry Kissos

Omer Meir

Lior Fritz

Guy Adam

Praveen Srinivasan

Mehmet Tek

An image is processed to determine semantic data about what is depicted in that image. The semantic data provides information about location in the image and semantic category of an object depicted within the image. The semantic categories may include 'background', “face”, “text”, “clothing”, “exercise”, and so forth. The semantic data may then be used to facilitate distribution of image data while limiting the information shared. For example, the semantic data may indicate the placement and position of a user exercising in first video data. Using the semantic data, second video data may be produced that obscures or replaces parts of the image that do not depict the user exercising. The second video data may then be used for providing feedback to the user, while preventing disclosure of other people, background, and so forth that are depicted in the first video data.

System for access control of image data using semantic data

A first user generates video data for performance of an activity, such as a fitness exercise, by performing the activity in front of a camera. Based on the video data, the amount of movement of different parts of the first user's body is determined. Data representing the position of the first user over time is generated. The data may take the form of a function or a signal that is based on the function. The locations of body parts that move significantly are prioritized over other body parts when determining this data. At a subsequent time, a second user performs the activity. The number of times the second user completes the activity is counted by determining the number of times the second user reaches a position corresponding to a maximum value in the data representing the position of the first user.

System for learning and tracking performance of activities

A system and method are disclosed for quickly and easily browsing to points of interest within a recorded video. In examples, the present system works in tandem with a segment list which includes a list of different video sequences of a stored video of an event. One typical example of a segment list is a play-by-play (PBP) which is prepared contemporaneously with a sporting event. Segments from a segment list may be associated with, or indexed to, corresponding sequences from the stored video. Thereafter, a user interface allows a user to interactively watch or browse the event by jumping to a desired point in the video for playback by selecting its associated segment from the user interface.

Browsing videos via a segment list

Images can be used to determine the size of an article of clothing matching the physical characteristics of a user. An image can include an article of clothing known to fit well or a representation of a body part, which can comprise the body part itself. Images can also include a reference item. Various techniques are used to determine one or more perspective distortions in the image based on analysis of the reference item and information known about the reference item. The one or more perspective distortions can be accounted for to determine physical characteristics of the article of clothing or the body part associated with the article of clothing. The physical characteristics can be used to determine the size of an article of clothing matching, or corresponding to, the physical characteristics.

Accurate size selection

Approaches provide for minimizing variations in the height of a camera of a computing device when estimating the distance to objects represented in image data captured by the camera. For example, a front-facing camera of a computing device can be used to capture a live camera view of a user. An application can analyze the image data to locate features of the user's face for purposes of aligning the user with the computing device. As the position and/orientation of the device changes with respect to the user, the image data can be analyzed to detect whether a location of a representation of a feature of the user aligns with the alignment element. Once the feature is aligned with the alignment element, a rear-facing camera (or other camera) can capture second image data of an object. The second image data can be analyzed to determine a geometric relationship between the rear-facing camera and the object, and the geometric relationship can be used to determine a distance to the object with respect to the computing device.

Minimizing variations in camera height to estimate distance to objects

Techniques to determine the three-dimensional structure of wearable items are disclosed. A scanning device can project waves at a wearable item and receive reflections of the waves. In embodiments, the scanning device projects light pulses and detects reflections of the light pulses. Depth measurements from the scanning device to the wearable item can be used to determine points corresponding to locations on the wearable item. A point cloud can be generated of the points, which can convey the three-dimensional structure of the wearable item. In embodiments, another wearable item can be determined to have a corresponding three-dimensional structure based on point cloud comparisons. The second wearable item can be suggested to a user as a wearable item having an appropriate fit for the user.

Computer vision techniques for generating and comparing three-dimensional point clouds

A video visual and audio query system for quickly identifying video within a large known corpus of videos being played on any screen or display. In one embodiment, the system can record via a mobile phone camera and microphone a live video clip from the TV and transcode it into a sequence of frame-signatures. The signatures representative of the clips can then be matched against the signatures of the TV content in a corpus across a network to identify the correct TV show or movie.

Video visual and audio query

An automated content recognition system accurately and reliably generates content identification information for multimedia content without accessing the multimedia content or a reliable source of the multimedia content. The system receives content-based queries having fingerprints of multimedia content. The system compares the individual queries to one another to match queries and thereby form query clusters that correspond to the same multimedia content. The system aggregates identification information from the queries in a cluster to generate reliable content identification information from otherwise unreliable identification information.

Automated multimedia content recognition

Technology described herein provides techniques for screen detection. This could be a dynamic screen, such as a display screen of an electronic device, or a static screen such as a street sign. In one aspect, a screen is detected in frames of image data. An active image and/or an average color image may be formed based on frames of image data. A set of lines that are candidate for boundaries of a screen are identified in the frames. The active image and/or an average color is integrated on both sides of each line in the set of lines to form integration results. A set of screen candidates are formed based on the integration results. Screens in the screen candidates are scored according to a criterion. One or top few of the screens in the set of screen candidates is selected based on the scoring.

Screen detector

Techniques for aligning images are disclosed. The frames might have been captured by a video camera on a hand held device, as one example. Collaboratively aligning related frames of image data is taught. Collaborative alignment determines a correspondence between pixels in pairs of the frames of image data, as well as a confidence in that correspondence. A coordinate system (or transformation) is assigned to each of the frames that is consistent with the correspondences between each of the frame pairs. The confidence in the respective correspondences may be used to provide a weighting to a correspondence when assigning the coordinate systems. Outlying frames may be discarded, and the process repeated for a robust solution.

Collaborative alignment of images

A method analyzes the visual content of media such as videos for collecting together visually-similar appearances in their constituent images (e.g. same scenes, same objects, faces of the same people.) As a result, the most relevant and salient (of clearest and largest presence) visual appearances depicted in the videos are presented to the user, both for the sake of summarizing the video content for the users to 'see before they watch' (that is, judge by the depicted video content in a filmstrip-like summary whether they want to mouse-click on the video and actually spend time watching it), as well as for allowing to users to further refine their video search result set according to the most relevant and salient video content returned (e.g. largest screen-time faces).

Apparatus and software system for and method of performing a visual-relevance-rank subsequent search

Systems and methods for identifying and searching video are disclosed. A video search and identification system includes a catalog representing relationships between video, data and/or objects to enable querying and search based on visual representations of video as well as data or other information associated with the video. In one example, the catalog includes video nodes, metadata nodes and card nodes, although additional or fewer node types may be used. A visual-based video identification system is provided to identify content in video sources. An unidentified video source is accessed and visual fingerprints of one or more frames are generated as query signatures for matching against a base set of known signatures. Confidence measures are generated at the bit level to assist in query signature processing.

Growing community of inventors

Los Gatos, CA, United States of America

Eran Borenstein