Amazon Technologies, Inc.

Google Inc.

Sharadh Ramaswamy

Amit Kumar Agrawal

Ambrish Tyagi

Jim Oommen Thomas

David Wayne Stafford

Volodymyr V Ivanchenko

David Foster Lieb

Matthew John Bridges

Stephen Vincent Mangiat

Madhur Khandelwal

Jingyu Cui

Kah Kuen Fu

Timothy Priestley Novikoff

Cheng-Hao Kuo

Juan Carlos Anorga

Evan Stephen Millar

Leslie Ikemoto

Shaun Tungseth

Austin Suszek

Jorge Verdu

Marc Cannon

Alexander Roe

Tianyang Ma

Sisil Sanjeev Mehta

Mugdha Kulkarni

Raja Ratna Murthy Ayyagari

Songbo Jin

Ruirui Jiang

Jeremy Selier

Gang Song

Ilya Vladimirovich Brailovskiy

Matthias Grundmann

Sourish Chaudhuri

Kenneth Conley

Joseph Edward Roth

Kenneth Mark Karakotsios

Isaac Scott Noble

Chang Yuan

Matthew Paul Bell

Oleg Rybakov

Geoffrey Heller

Jason Robert Weber

Dmitri Khoklov

K R Suveetha Kamatchi

Implementations described herein relate to methods, devices, and computer-readable media to perform gating for video analysis. In some implementations, a computer-implemented method includes obtaining a video comprising a plurality of frames and corresponding audio. The method further includes performing sampling to select a subset of the plurality of frames based on a target frame rate and extracting a respective audio spectrogram for each frame in the subset of the plurality of frames. The method further includes reducing resolution of the subset of the plurality of frames. The method further includes applying a machine-learning based gating model to the subset of the plurality of frames and corresponding audio spectrograms and obtaining, as output of the gating model, an indication of whether to analyze the video to add one or more video annotations.

Gating model for video analysis

Implementations relate to causing a command to be executed based on an image. In some implementations, a computer-implemented method includes obtaining and programmatically analyzing an image to determine suggested actions. The method causes a user interface to be displayed that includes user interface elements corresponding to default actions, and to suggested actions that are determined based on analyzing the image. The method receives user input indicative of selection of a particular action from the default actions and the suggested actions. The method causes a command to be executed by a computing device for the particular action that was selected.

Suggested actions for images

A computer-implemented method includes determining interesting moments in a video. The method further includes generating video segments based on the interesting moments, wherein each of the video segments includes at least one of the interesting moments from the video. The method further includes generating a collage from the video segments, where the collage includes at least two windows and wherein each window includes one of the video segments.

Collage of interesting moments in a video

Implementations relate to removal of one or more images from a view of a plurality of images. In some implementations, a method includes obtaining a plurality of images, programmatically analyzing the plurality of images to determine a plurality of image features, and determining one or more image categories for the plurality of images based on the image features. The method further includes identifying a subset of the plurality of images based on the image categories, wherein each image of the subset is associated with an image category for archival. The method further includes causing a user interface to be displayed that includes one or more images of the subset, receiving user input to archive at least one of the one or more images, and in response to the user input, removing the at least one of the images from a view of the plurality of images.

Image archival based on image categories

A method for compression across multiple images is provided. The method includes receiving a plurality of images at a storage device. For each image of the plurality of images, the method includes generating codewords using lossy compression, each codeword defined as a bit string. The method further includes generating an entropy coding table based on the codewords generated for each image of the plurality of images and entropy encoding the plurality of images using the generated entropy coding table.

Compression across multiple images

Systems and approaches provide for user interfaces that are based on object tracking. For example, the object may be a user's head or face. As the user moves his head or face and/or tilts a computing device, the content displayed on the computing device will adapt to the user's perspective. The content may include three-dimensional (3D) graphical elements projected onto a two-dimensional (2D) plane and/or the graphical elements can be associated with textural shading, shadowing or reflections that change according to user or device motion to give the user the impression that the user is interacting with the graphical elements in 3D environment. A state of motion of the device can be determined and jitter and/or latency corresponding to the rendering of content can be altered so as to minimize or decrease jitter when the device is stationary and/or to decrease or minimize latency when the device is in motion.

Dynamic object tracking for user interfaces

Various embodiments enable a primary user to be identified and tracked using stereo association and multiple tracking algorithms. For example, a face detection algorithm can be run on each image captured by a respective camera independently. Stereo association can be performed to match faces between cameras. If the faces are matched and a primary user is determined, a face pair is created and used as the first data point in memory for initializing object tracking. Further, features of a user's face can be extracted and the change in position of these features between images can determine what tracking method will be used for that particular frame.

Primary user selection for head tracking

Approaches are described for utilizing a combination of location determining and tracking techniques to attempt to provide driver assistance under a variety of different circumstances, such as location assistance to locate or get back to a particular object or location, and visual and/or audible assistance to alert a driver of potential danger. For example, a driver assistance system can be provided that can monitor traffic conditions as well as attentiveness of a driver using a portable computing device (such as a mobile phone) to provide visual and/or audible notifications to alert the driver to potential dangers and/or take defensive action. The driver assistance system can further be used to automatically determine the location of a user's parked automobile and provide assistance, such as directions, to get back to the location where the user parked his or her automobile.

Driver assist using smart mobile devices

An electronic device can have two or more pairs of cameras capable of performing three-dimensional imaging. In order to provide accurate disparity information, these cameras should be sufficiently calibrated. Automatic calibration can be performed by periodically capturing images with a pair of front-facing cameras and locating matching facial or other feature points in corresponding images captured by those cameras. Correspondences can be detected between feature points and the corresponding feature points can be normalized and outlier feature points can be rejected. A transformation matrix can be determined using at least a portion of remaining feature points and can be used to determine rotation and translation parameters to correct for misalignment between the cameras. The calibration parameters can be refined or otherwise adjusted, and can be used or stored for use in correcting images subsequently captured by those cameras.

Online calibration of cameras

Various embodiments provide a method for determining conditions associated with the presence of a user in order to perform object detection. For example, various heuristic tests can be used to identify movement within a sequence of images or to identify user engagement therewith and, if movement or user engagement is identified, stereo disparity can be computed in an attempt to detect an object within a predetermined distance of the computing device. Accordingly, if the distance of the object is within the predetermined distance, object detection is performed. If the object is beyond the threshold distance, however, object detection is not performed in order to conserve battery life of the computing device.

Power saving approaches to object detection

Embodiments enable detecting an occurrence such as an impact event, which can occur when an electronic device impacts a surface or object (e.g., from a fall), which may have occurred while the device was powered off, in a low power mode, or other state where the device cannot detect such an impact using an accelerometer or other such sensor. Various calibration processes can be used to determine an amount of misalignment between two or more cameras of the device, where an amount of misalignment more than an allowable threshold can be indicative of the impact event, at which point one or more system checks can be performed to determine whether the device components (e.g., memory, hard-disk, drives, antennas, etc.) or certain portions or components of the device are operating properly.

Detecting impact events for dropped devices

Approaches are described which enable a computing device (e.g., mobile phone, tablet computer) to utilize one or more facial recognition techniques to control access to the device and to detect when artificial representations of a user, such as a picture or photograph, are being used in an attempt to gain access to the device. Evidence indicative of artificial representations may include lack of changes in facial skin color between multiple images captured by a camera, ability to track one or more features of the human face while the camera is rotated or moved, presence of secular reflections caused by an illumination device, absence of shadows in the image, and others.

Fraud detection for facial recognition systems

Growing community of inventors

Sunnyvale, CA, United States of America

Sharadh Ramaswamy