SYSTEM, METHOD AND APPARATUS FOR PERFORMING FACIAL RECOGNITION

• US 20150317511A1
• Filed: 06/24/2015
• Published: 11/05/2015
• Est. Priority Date: 11/07/2013
• Status: Abandoned Application

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Orbeus, Inc. (Amazon.com, Inc.)
• Inventors: Li, Yi, Liu, Tianqiang, Chen, Hao, Wang, Meng
• Application Number:
  US14/749,465
• Publication Number:
  US 20150317511A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/583 using metadata automatically derived from the content
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 16/955 using information identifiers, e.g. uniform resource locators [URL]
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06V 40/165 using facial parts and geometric relationships
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A method for recognizing a face in an image is performed by a facial recognition system. The system retrieves an image and detects a face within the image. The system then determines a set of facial feature positions for a set of facial features. The set of facial feature positions are used to separate the face into a set of facial feature parts. For each part, the system extracts a set of image features. The extracted features are concatenated into a full feature. The system performs dimension reduction on the full feature to derive a final feature.

Claims

1. An image processing system for performing facial recognition on an image, the image processing system comprising:

a first server computer, the first server computer including a processor, a network interface coupled to the processor and operatively coupled to a database, and a server software application operating on the processor;

ii) the server software application adapted to;

1) over the network interface, retrieve an image from a client computer or a second server computer;

2) retrieve a set of models from the database;

3) extract a final feature from the image;

4) apply each model in the set of models on the final feature to generate a set of comparison scores;

5) select the highest score from the set of comparison scores;

6) where the highest score is below a threshold, indicate that a face is not recognized in the image;

7) where the highest score is over the threshold, associate the image with a label, wherein the label is associated with the model corresponding to the highest score, thereby forming an association; and

8) store the association into the database.

2. The image processing system of claim 1, wherein the server software application is further adapted to determine that the image correctly labeled.

3. The image processing system of claim 1, wherein the server software application is further adapted to:

i) retrieve a set of training images over the network interface;

ii) for each training image in the set, extract a training image final feature; and

iii) perform model training on the set of training image final features to derive a recognition model, wherein the recognition model is a model in the set of models.

4. The image processing system of claim 1, wherein the server software application is further adapted to:

i) retrieve a video dip over the network interface;

ii) select a set of representing frames from the video clip;

iii) perform facial recognition on each frame in the set of representing frames to identify a second face, thereby forming a set of second faces; and

from the set of second faces, select one second face as an identified face for the video clip.

5. The image processing system of claim 1, wherein the second server computer is a social media networking server.