Google Inc.

University of South Florida

Government of the United States of America, As Represented by the Secretary of Commerce

Pranab Mohanty

Ullas Gargi

Sudeep Sarkar

Rangachar Kasturi

P Jonathon Phillips

A video hosting web site receives uploaded video content and processes the video to determine ad surfaces. The ad surfaces comprise spatio-temporal regions of the video suitable for placement of advertisement such as background surfaces or other regions of low interest. The uploaded video and ad surfaces are stored to a video database that is accessible to viewers visiting the video hosting web site. When a shared video is requested, the video hosting site provides the requested video content together with the ad surfaces and advertisements. The advertisements are blended with the ad surfaces in the video at playtime so that the advertisements appear as part of the video scene. This allows the video hosting web site to present advertisements to the viewer without significantly distracting the viewer from the requested content.

Finding and populating spatial ad surfaces in video

A novel, linear modeling method to model a face recognition algorithm based on the match scores produced by the algorithm. Starting with a distance matrix representing the pair-wise match scores between face images, an iterative stress minimization algorithm is used to obtain an embedding of the distance matrix in a low-dimensional space. A linear transformation used to project new face images into the model space is divided into two sub-transformations: a rigid transformation of face images obtained through principal component analysis of face images and a non-rigid transformation responsible for preserving pair-wise distance relationships between face images. Also provided is a linear indexing method using the linear modeling method to perform the binning or algorithm-specific indexing task with little overhead.

Indexing face templates using linear models

A method of reconstructing biometric face image templates of a face recognition system (FRS) using the match scores or distances provided by the FRS. The match scores represent the distance between a image introduced to the FRS and the unknown image template stored in the FRS. The present method uses an affine transformation approximating the unknown algorithm within the FRS and the match scores provided by the FRS to determine the coordinates of the unknown target template. The coordinates of the unknown target template are then applied to a pseudo-inversion of the affine transformation to produce a reconstructed image template of the unknown target. This reconstructed image template can then be used to 'break-in' to the FRS.

Growing community of inventors

Tampa, FL, United States of America

Pranab Mohanty