Techniques for enabling or establishing the use of face recognition algorithms

• US 8,385,633 B2
• Filed: 12/07/2010
• Issued: 02/26/2013
• Est. Priority Date: 03/12/2006
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Gokturk, Salih Burak, Anguelov, Dragomir, Torresani, Lorenzo, Vanhoucke, Vincent, Shah, Munjal, Vu, Diem, Lee, Kuang-Chih
• Primary Examiner(s): Newman, Michael A
• Application Number:
  US12/961,790
• Publication Number:
  US 20110075919A1
• Time in Patent Office: 812 Days
• Field of Search: 382/159, 382/155, 382/224, 382/225, 382/309, 382/118, 382/115, 382/154, 382/285, 382/300
• US Class Current: 382/159
• CPC Class Codes: G06V 10/7557 based on appearance, e.g. active appearance models [AAM]
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

Embodiments described herein facilitate or enhance the implementation of image recognition processes which can perform recognition on images to identify objects and/or faces by class or by people.

Claims

1. A computer-implemented method comprising:

programmatically selecting, from a collection of images, a set of candidate images for a training set;

providing the set of candidate images to a user for input;

designating images from the set of candidate images for the training set based on input from the user;

performing a recognition of an input image;

wherein performing the recognition includes comparing the input image with one or more images of the training set, and determining, based at least in part on comparing the input image with the one or more images of the training set, that a confidence value of the recognition satisfies a defined threshold;

providing a results of the recognition to the user, wherein the result includes a plurality of images that have a recognized object or face that is represented in an image of the training set;

determining an error rate base on user feedback regarding the results, wherein the error rate is associated with a measurement of instances of erroneous recognition of the input image from the plurality of images provided with the result; and

in response to the user feedback, adjusting the defined threshold based on the error rate.

2. The computer-implemented method of claim 1, wherein programmatically selecting includes identifying a cluster of images from a collection, wherein each image in the cluster is determined to contain an image of an object or person.

3. The computer-implemented method of claim 2, wherein identifying a cluster of images includes determining cluster size and compactness.

4. The computer-implemented method of claim 1, wherein providing a result of the recognition to the user includes providing to the user a plurality of recognized images that each include a recognized object or face that is contained in an image of the training set, and wherein the method further comprises enabling the user to identify if any of the images in the plurality of recognized images includes an erroneous recognition.

5. The computer-implemented method of claim 1, wherein adjusting the defined threshold comprises reducing or increasing the defined threshold based on the measurement of the instances of erroneous recognition.

6. The computer-implemented method of claim 1, wherein programmatically selecting includes performing a simulated recognition of images in a collection.

7. A non-transitory computer-readable storage medium having stored therein instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

selecting, from a collection of images, a set of candidate images for a training set;

providing the set of candidate images to a user for input;

designating images from the set of candidate images for the training set based on input from the user;

performing a recognition of an input image,

wherein performing the recognition includes comparing the input image with one or more images of the training set, and determining, based at least in part on comparing the input image with the one or more images of the training set, that a confidence value of the recognition satisfies a defined threshold;

providing a result of the recognition to the user, wherein the result includes a plurality of images that have a recognized object or face that is represented in an image of the training set;

determining an error rate based on user feedback regarding the result, wherein the error rate is associated with a measurement of instances of erroneous recognition of the input image from the plurality of images provided with the result; and

in response to the user feedback, adjusting the defined threshold based on the error rate.

8. The non-transitory computer-readable storage medium of claim 7, wherein instructions for selecting includes instructions for identifying a cluster of images from a collection, wherein each image in the cluster is determined to contain an image of an object or person.

9. The non-transitory computer-readable storage medium of claim 8, instructions for identifying a cluster of images includes instructions for determining cluster size and compactness.

10. The non-transitory computer-readable storage medium of claim 7, wherein instructions for providing a result of the recognition to the user includes instructions for providing to the user a plurality of recognized images that each include a recognized object or face that is contained in an image of the training set, and wherein the non-transitory computer-readable storage medium further comprises instructions for enabling the user to identify if any of the images in the plurality of recognized images includes an erroneous recognition.

11. The non-transitory computer-readable storage medium of claim 7, wherein instructions for adjusting the defined threshold comprises instructions for reducing or increasing the defined threshold based on the measurement of instances of erroneous recognition.

12. The non-transitory computer-readable storage medium of claim 7, wherein instructions for selecting includes instructions for performing a simulated recognition of images in a collection.

13. A computer system comprising:

a memory that stores a set of instructions;

one or more processors that use instructions from the set of instructions to;

select, from a collection of images, a set of candidate images for a training set;

provide the set of candidate images to a user for input;

designate images from the set of candidate images for the training set based on input from the user;

perform a recognition of an input image, including comparing the input image with one or more images of the training set, and determining, based at least in part on comparing the input image with the one or more images of the training set, that a confidence value of the recognition satisfies a defined threshold;

provide a result of the recognition to the user, wherein the result includes a plurality of images that have a recognized object or face that is represented in an image of the training set;

determine an error rate based on user feedback regarding the result, wherein the error rate is associated with a measurement of instances of erroneous recognition of the input image from the plurality of images provided with the result; and

in response to the user feedback, adjust the defined threshold based on the error rate.

14. The computer system of claim 13, wherein the one or more processors select the set of candidate images by identifying a cluster of images from a collection, wherein each image in the cluster is determined to contain an image of an object or person.

15. The computer system of claim of claim 14, wherein the one or more processors identify the cluster of images by determining cluster size and compactness.

16. The computer system of claim 13, wherein the one or more processors provide the result of the recognition to the user by providing plurality of recognized images that each include a recognized object or face that is contained in an image of the training set, and wherein the one or more processors enable the user to identify if any of the images in the plurality of recognized images includes an erroneous recognition.

17. The computer system of claim 13, wherein the one or more processors adjust the defined threshold by reducing or increasing the defined threshold based on the measurement of instances of erroneous recognition.