Google researchers See Jung Ryu and Florian Schroff have developed an "electronic screen protector" application that can immediately spot people glancing at a user's handheld screen.
They have created a video demo indicating they have installed a lightweight machine-learning model on a Pixel smartphone that employs its front-facing camera for fast gaze detection. When the user holds the smartphone up to chat or watch a private video, the algorithm will detect when a person from behind starts looking at the screen as well, detecting gaze within two milliseconds and recognizing a face in 47 milliseconds.
The demo implies the use of FaceNet, a facial-recognition neural network, and the GazeNet gaze-estimation neural network.
Ryu and Schroff say their invention delivers enhanced privacy when using big-screen smartphones in public locales, building on face-authentication systems already on smartphones by using the enrolled face to differentiate between users and strangers, while the gaze-estimation model helps identify the stranger's gaze almost instantaneously.
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA