Google Inc.

Gautam Prasad

Ting Liu

Bohyung Han

Phuc Xuan Nguyen

Dmitry Lagun

Florian Schroff

A method includes determining a measured eye gaze position of an eye of a user. The method also includes determining a first incremental change in the measured eye gaze position by processing the measured eye gaze position by a long short-term memory (LSTM) model, and determining a first predicted eye gaze position of the eye at a first future time based on the measured eye gaze position and the first incremental change. The method additionally includes determining a second incremental change in the first predicted eye gaze position by processing the first predicted eye gaze position by the LSTM model, and determining a second predicted eye gaze position of the eye at a second future time subsequent to the first future time based on the first predicted eye gaze position and the second incremental change.

Machine learning based forecasting of human gaze

Systems and methods for a weakly supervised action localization model are provided. Example models according to example aspects of the present disclosure can localize and/or classify actions in untrimmed videos using machine-learned models, such as convolutional neural networks. The example models can predict temporal intervals of human actions given video-level class labels with no requirement of temporal localization information of actions. The example models can recognize actions and identify a sparse set of keyframes associated with actions through adaptive temporal pooling of video frames, wherein the loss function of the model is composed of a classification error and a sparsity of frame selection. Following action recognition with sparse keyframe attention, temporal proposals for action can be extracted using temporal class activation mappings, and final time intervals can be estimated corresponding to target actions.

Growing community of inventors

Los Angeles, CA, United States of America

Gautam Prasad