Google Inc.

Pannag Raghunath Sanketi

Rodrigo Lima Carceroni

Hrishikesh Balkrishna Aradhye

Daniel Ramage

Amarnag Subramanya

Ruei-Sung Lin

Wei Hua

Marius Renn

Wolfgang Grieskamp

Derya Ozkan

Suril Shah

Brett Rolston Lider

Wei Chai

Soroosh Mariooryad

Jamie Gennis

Peter Wilhelm Ludwig

Wesley Robert Tarle

Shiyu Hu

Seyed Mojtaba Seyedhosseini Tarzjani

Ibrahim Elbouchikhi

Sachin Kotwani

Provided is an application development platform and associated software development kits ('SDKs') that provide comprehensive services for generation, deployment, and management of machine-learned models used by computer applications such as, for example, mobile applications executed by a mobile computing device. In particular, the application development platform and SDKs can provide or otherwise leverage a unified, cross-platform application programming interface (“API”) that enables access to all of the different machine learning services needed for full machine learning functionality within the application. In such fashion, developers can have access to a single SDK for all machine learning services.

Application development platform and software development kits that provide comprehensive machine learning services

The present disclosure provides systems and methods for on-device machine learning. In particular, the present disclosure is directed to an on-device machine learning platform and associated techniques that enable on-device prediction, training, example collection, and/or other machine learning tasks or functionality. The on-device machine learning platform can include a context provider that securely injects context features into collected training examples and/or client-provided input data used to generate predictions/inferences. Thus, the on-device machine learning platform can enable centralized training example collection, model training, and usage of machine-learned models as a service to applications or other clients.

On-device machine learning platform

In one example, a method includes determining, by a processor () of a wearable computing device () and based on motion data generated by a motion sensor () of the wearable computing device, one or more strokes. In this example, the method also includes generating, by the processor and based on the motion data, a respective attribute vector for each respective stroke from the one or more strokes and classifying, by the processor and based on the respective attribute vector, each respective stroke from the one or more strokes into at least one category. In this example, the method also includes determining, by the processor and based on a gesture library and the at least one category for each stroke from the one or more strokes, a gesture. In this example, the method also includes performing, by the wearable device and based on the gesture, an action.

Identifying gestures using motion data

An example method includes displaying, by a display () of a wearable device (), a content card (B); receiving, by the wearable device, motion data generated by a motion sensor () of the wearable device that represents motion of a forearm of a user of the wearable device; responsive to determining, based on the motion data, that the user has performed a movement that includes a supination of the forearm followed by a pronation of the forearm at an acceleration that is less than an acceleration of the supination, displaying, by the display, a next content card (C); and responsive to determining, based on the motion data, that the user has performed a movement that includes a supination of the forearm followed by a pronation of the forearm at an acceleration that is greater than an acceleration of the supination, displaying, by the display, a previous content card (A).

Touchless user interface navigation using gestures

A computer-implemented technique can receive a plurality of photos and automatically select a subset of the plurality of photos having a high degree of representativeness by jointly maximizing both photo quality and photo diversity to obtain a photo album. The technique can determine one or more clusters for the photo album using a hierarchical clustering algorithm, and store the photo album according to the one or more clusters. The technique can control the manner in which the photo album is displayed using the one or more clusters. The technique can adjust at least one of the one or more clusters and the automatic photo album generation based on user input. The user input can include at least one of adding, deleting, and moving a photo with respect to the one or more clusters. The technique can then re-cluster, automatically generate a new photo album, and/or adjust the presentation.

Techniques for user customization in a photo management system

The disclosed technology includes techniques for improved content coverage in automatically-generated content summaries. The technique may include clustering a set of input content, determining diffusion for each cluster, and selecting representatives of each cluster to optimize other secondary metrics. Various types of input content may be used, including groups of images, video clips, or other multimedia content. Contiguous content may be manually or programmatically divided into discrete portions before clustering, for example, a lengthy video divided into a number of short clips. In some implementations, the disclosed technique may be implemented effectively on a mobile device. In other words, the processing required may be computationally feasible for execution on a smartphone or similar device.

Systems and methods for improved coverage of input media in content summarization

Techniques for determining motion saliency in video content using center-surround receptive fields. In some implementations, images or frames from a video may be apportioned into non-overlapped regions, for example, by applying a rectilinear grid. For each grid region, or cell, motion consistency may be measured between the center and surround area of that cell across frames of the video. Consistent motion across the center-surround area may indicate that the corresponding region has low variation. The larger the difference between center-surround motions in a cell, the more likely the region has high motion saliency.

Systems and methods for determining motion saliency

A program module executing in a first process space of a mobile computing device receives a buffer request from a graphics driver running in a second process space of the mobile computing device, wherein the second process space is isolated from the first process space. The program module assigns a buffer to the graphics driver to store image data processed by a graphical processing unit (GPU) controlled by the graphics driver. The program module receives a release of the buffer from the graphics driver. The program module assigns the buffer to a media encoder driver for a hardware media encoder to encode the image data in the buffer into a file.

Method and system for sharing a buffer between a graphics processing unit and a media encoder

Methods and apparatus for suggesting image, video, and image album titles are presented. A machine-learning service executing on a mobile platform receives feature-related data. The feature-related data includes image-related data related to one or more images received from an application executing on the mobile platform and platform-related data received from the mobile platform. The image-related data and the platform-related data differ. The machine-learning service generates a title related to the one or more images by performing a machine-learning operation on the feature-related data. The machine-learning service sends the title related to the one or more images to the application.

Growing community of inventors

Mountain View, CA, United States of America

Pannag Raghunath Sanketi