Face template balancing

Patent metadata

Field	Value
Publication number	US-9779285-B2
Application number	US-201615287792-A
Country	US
Kind code	B2
Filing date	Oct 7, 2016
Priority date	Jul 19, 2013
Publication date	Oct 3, 2017
Grant date	Oct 3, 2017

Abstract

Implementations generally relate to face template balancing. In some implementations, a method includes generating face templates corresponding to respective images. The method also includes matching the images to a user based on the face templates. The method also includes receiving a determination that one or more matched images are mismatched images. The method also includes flagging one or more face templates corresponding to the one or more mismatched images as negative face templates.

First claim

What is claimed is:

1. A computer-implemented method comprising:
generating a target face template corresponding to a target face in a target image, wherein the target face template includes at least one indication of at least one facial feature of the target face;
retrieving from a storage device, a face model for a face of a user, wherein the face model includes at least one model indication of one or more model facial features of the face of the user, and wherein the at least one model indication is associated with reference facial features of a plurality of positive face templates generated from a first set of images previously verified to depict the user;
determining a match of the at least one indication of the at least one facial feature of the target face template with the at least one model indication;
in response to determining the match, associating a tag with the target image, wherein the tag identifies the target face as the user;
retrieving from the storage device, one or more negative face templates generated from a second set of images having one or more faces previously verified to not be the user, wherein the one or more negative face templates include at least one indication of one or more differentiating facial features;
determining whether the tag correctly identifies the target face as the user, based on comparing the at least one indication of the at least one facial feature of the target face template with the at least one indication of the one or more differentiating facial features of the one or more negative face templates; and
in response to determining the tag correctly identifies the target face as the user, transmitting the target image and the tag to a computing device through a data communication network.

2. The method of claim 1 , further comprising,
providing to a device, one or more test images;
receiving, by the device, user input that indicates that one or more images of the one or more test images are one or more mismatched images that do not depict the user; and
designating one or more face templates that correspond to the one or more mismatched images as the one or more negative face templates.

3. The method of claim 1 , wherein determining whether the tag correctly identifies the target face as the user includes determining that the target image does not depict the user based on the target face template based on a level of similarity between the target face template and at least one of the one or more negative face templates being greater than a level of similarity between the target face template and the face model.

4. The method of claim 1 , wherein determining whether the tag correctly identifies the target face as the user includes determining that the target image depicts the user based on the target face template corresponding to the target image based on a level of similarity between the target face template and at least one of the one or more negative face templates being less than a level of similarity between the target face template and the face model.

5. The method of claim 4 , further comprising, in response to determining that the tag correctly identifies the target face as the user, updating the face model with the target face template corresponding to the target image.

6. The method of claim 1 , further comprising:
receiving, from a device, user input that indicates that the target image is a mismatched image that does not depict the user.

7. The method of claim 1 , further comprising:
receiving, from a device, user input that indicates that the target image is a correctly matched image that depicts the user.

8. The method of claim 1 , wherein determining whether the tag correctly identifies the target face as the user includes:
determining the one or more differentiating characteristics that differentiate the one or more negative face templates from the plurality of positive face templates; and
determining that the target face template includes at least one differentiating characteristic of the one or more differentiating characteristics indicating that the target image does not depict the user.

9. The method of claim 8 wherein the one or more differentiating characteristics include at least one of:
one or more facial features in the one or more negative face templates or in the plurality of positive face templates, and
one or more colors provided in the face described by the facial features.

10. The method of claim 1 ,
wherein determining whether the tag correctly identifies the target face as the user includes providing a negative weight for the determining based on the one or more differentiating facial features in the target image that do not match the face of the user.

11. The method of claim 1 , wherein the target face template corresponding to the target image describes at least one of:
a location of the target face within the target image, and
one or more distances between a plurality of facial features of the face.

12. A system comprising:
one or more processors; and
logic encoded in one or more non-transitory tangible media for execution by the one or more processors and when executed operable to perform operations comprising:
generating a plurality of face templates corresponding to respective images, wherein each face template of the plurality of face templates includes a description of each of facial features of a face depicted in a corresponding one of the respective images and a value for each of the facial features;
matching the respective images to a user identity of a user by determining similarities of at least one of the facial features to one or more reference facial features in one or more reference images that depict the user;
determining that at least one image of the respective images is at least one mismatched image that does not depict the user associated with the user identity, wherein the at least one mismatched image includes one more differentiating facial features, and wherein the determining is based on first received user input;
designating at least one face template corresponding to the at least one mismatched image as at least one negative face template, wherein the at least one face template includes a value for each of the one or more differentiating facial features;
generating a face model for the user based on combining the values of the facial features of one or more face templates of the plurality face templates that are not the at least one negative face template;
using the face model to determine whether a target image is matched to the user identity of the user associated with the face model;
verifying whether the target image is matched to the user identity by determining whether the target image includes the one or more differentiating facial features; and
in response to verifying the target image is matched to the user identity, transmitting the target image and at least one tag associated with the user identity, to a computing device through a data communication network.

13. The system of claim 12 , wherein the logic when executed by the one or more processors is operable to perform further operation comprising:
determining, based on second received user input, that one or more of the matched respective images are one or more correctly matched images that depict the user of the user identity; and
designating one or more face templates corresponding to the one or more correctly matched images as one or more positive face templates, wherein the face model is generated based on the one or more positive face templates.

14. The system of claim 12 , wherein the operation of using the at least one

Assignees

• Google Inc

Inventors

• Mcphie Jonathan
• Adam Hartwig
• Fredinburg Dan
• Masterov Alexei

Classifications

• G06V10/7788 Primary

  the supervisor being a human, e.g. interactive learning with a human teacher · CPC title

• G06V10/772

  Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries · CPC title

• G06V40/164

  using holistic features · CPC title

• G06F18/28

  Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries · CPC title

• G06F18/41

  Interactive pattern learning with a human teacher · CPC title