Face recognition and image search system using sparse feature vectors, compact binary vectors, and sub-linear search

Abstract

Systems and methods are provided for performing automated face recognition and comparison. An input image of a face may be received and cropped, and the image may be processed through a deep neural network (DNN) to produce a k-dimensional feature vector. The k-dimensional feature vector may be converted to a k-dimensional binary vector by transforming each value in the vector to either 1 or 0. To search for nearest matches of the image in a database of gallery images of faces, the system may compare sub-strings of the binary vector to hash tables created from sub-strings of the gallery images, enabling sub-linear searching that allows locating the closest matches from among the entire gallery without requiring an exhaustive linear search of the entire gallery.

Description

REFERENCE TO RELATED APPLICATIONS[0001]This application is a continuation of U.S. application Ser. No. 15 / 724,936, filed Oct. 4, 2017, which claims the benefit of U.S. Provisional Application No. 62 / 405,721, filed Oct. 7, 2016, the entire contents of each of which are incorporated herein by reference.FIELD OF THE INVENTION[0002]This relates to systems and methods for face recognition and image searching.BACKGROUND OF THE INVENTION[0003]Images and videos are being disseminated in the open source, particularly on the internet, at an unprecedented rate. It is estimated that on average, every minute, hundreds of thousands of images are shared on social media websites alone. On YouTube, on average, over 100 hours of video comprising over 8 million images are shared every minute. This vast number of images can contain information that is highly valuable. For example, the ability to perform face recognition across the internet could be useful in finding exploited children, protecting or ex...

AI Technical Summary

Benefits of technology

The patent describes a system for quickly processing and analyzing millions of images of faces. The system uses a deep neural network to extract information from the images and compare it to a database of faces. Information is then extracted and used to create a more manageable binary vector, which can be quickly compared to other images using sub-linear search methods. This system can also ingest images from websites and extract information from them using the DNN. The invention allows for efficient and accurate processing of image data, which can be used for various applications such as facial recognition.

Problems solved by technology

In the field of automated face recognition, known techniques leverage human-designed algorithms to create representations of images of faces that are inefficient and / or slow to create, unreliable in their ability to differentiate faces, and too large to efficiently or feasibly be compared to large numbers of other such representations of images of faces.

Images