Image recognition method and apparatus, device, and computer storage medium

Patent metadata

Field	Value
Publication number	US-11687779-B2
Application number	US-202117208611-A
Country	US
Kind code	B2
Filing date	Mar 22, 2021
Priority date	Jun 29, 2020
Publication date	Jun 27, 2023
Grant date	Jun 27, 2023

Abstract

An image recognition method is provided, which is related to a technical field of artificial intelligence, and in particular, to a technical field of image processing. An implementation includes: performing five-sense-organ recognition on a preprocessed human face image and marking positions of the human facial five sense organs in the human face image, to obtain the marked human face image; determining human face images at multiple scales of the marked human face image, inputting the human face images of multiple scales into a backbone network model, and performing feature extraction, to obtain a wrinkle feature of the human face image at each of the multiple scales; and fusing the wrinkle feature at each scale that is located in a same area of the human face image, to obtain a wrinkle recognition result of the human face image.

First claim

What is claimed is:

1. An image recognition method, comprising:
performing five-sense-organ recognition on a preprocessed human face image and marking positions of the human facial five sense organs in the human face image, to obtain a marked human face image;
determining human face images at multiple scales of the marked human face image, inputting the human face images of multiple scales into a backbone network model, and performing feature extraction, to obtain a wrinkle feature of the human face image at each of the multiple scales; and
fusing the wrinkle feature at each scale that is located in a same area of the human face image, to obtain a wrinkle recognition result of the human face image.

2. The method according to claim 1 , further comprising:
performing homogenization processing on pixels of the human face image, to obtain a homogenized human face image; and
performing normalization processing on pixel variances of the homogenized human face image, to obtain the preprocessed human face image.

3. The method according to claim 1 , wherein the inputting the marked human face image into a backbone network model and performing feature extraction, to obtain wrinkle features of the human face image at different scales, comprises:
performing multi-scale stretching and retraction processing on the marked human face image, to obtain the human face images of multiple scales; and
inputting the human face images of multiple scales into the backbone network model, to obtain a wrinkle feature of the human face image at each of the multiple scales.

4. The method according to claim 3 , further comprising:
inputting the human face images of multiple scales into the backbone network model, to obtain a recognition result of areas of the human facial five sense organs.

5. The method according to claim 3 , wherein the fusing the wrinkle features at different scales that are located in a same area of the human face image, to obtain a wrinkle recognition result of the human face image, comprises:
ignoring the wrinkle recognition result of the human face image in a case where a face area corresponding to the wrinkle recognition result of the human face image is an area of the five sense organs that has no wrinkle.

6. An image recognition apparatus, comprising:
at least one processor; and
a memory communicatively connected to the at least one processor, wherein
the memory stores instructions executable by the at least one processor, the instructions are executed by the at least one processor to enable the at least one processor to:
perform five-sense-organ recognition on a preprocessed human face image and mark positions of the human facial five sense organs in the human face image, to obtain a marked human face image;
determine human face images at different scales of the marked human face image, input the human face images of multiple scales into a backbone network model, and performing feature extraction, to obtain a wrinkle feature of the human face image at each of the multiple scales; and
fuse the wrinkle feature at each scale that is located in a same area of the human face image, to obtain a wrinkle recognition result of the human face image.

7. The apparatus according to claim 6 , wherein the instructions are executed by the at least one processor to further enable the at least one processor to:
perform homogenization processing on pixels of the human face image, to obtain a homogenized human face image; and
perform normalization processing on pixel variances of the homogenized human face image, to obtain the preprocessed human face image.

8. The apparatus according to claim 6 , wherein the instructions are executed by the at least one processor to further enable the at least one processor to:
perform multi-scale stretching and retraction processing on the marked human face image, to obtain the human face images of multiple scales; and
input the human face images of multiple scales into the backbone network model to obtain a wrinkle feature of the human face image at each of the multiple scales.

9. The apparatus according to claim 8 , wherein the instructions are executed by the at least one processor to further enable the at least one processor to input the human face images of multiple scales into the backbone network model, to obtain a recognition result of areas of the human facial five sense organs.

10. The apparatus according to claim 8 , wherein the instructions are executed by the at least one processor to further enable the at least one processor to ignore the wrinkle recognition result of the human face image in a case where a face area corresponding to the wrinkle recognition result of the human face image is an area of the five sense organs that has no wrinkle.

11. A non-transitory computer-readable storage medium for storing computer instructions, wherein the computer instructions, when executed by a computer, cause the computer to:
perform five-sense-organ recognition on a preprocessed human face image and mark positions of the human facial five sense organs in the human face image, to obtain a marked human face image;
determine human face images at multiple scales of the marked human face image, input the human face images of multiple scales into a backbone network model, and perform feature extraction, to obtain a wrinkle feature of the human face image at each of the multiple scales; and
fuse the wrinkle feature at each scale that is located in a same area of the human face image, to obtain a wrinkle recognition result of the human face image.

12. The non-transitory computer-readable storage medium according to claim 11 , wherein the computer instructions, when executed by a computer, further cause the computer to:
perform homogenization processing on pixels of the human face image, to obtain a homogenized human face image; and
perform normalization processing on pixel variances of the homogenized human face image, to obtain the preprocessed human face image.

13. The non-transitory computer-readable storage medium according to claim 11 , wherein the computer instructions, when executed by a computer, further cause the computer to:
perform multi-scale stretching and retraction processing on the marked human face image, to obtain the human face images of multiple scales; and
input the human face images of multiple scales into the backbone network model, to obtain a wrinkle feature of the human face image at each of the multiple scales.

14. The non-transitory computer-readable storage medium according to claim 13 , wherein the computer instructions, when executed by a computer, further cause the computer to:
input the human face images of multiple scales into the backbone network model, to obtain a recognition result of areas of the human facial five sense organs.

15. The non-transitory computer-readable storage medium according to claim 13 , wherein the computer instructions, when executed by a computer, further cause the computer to:
ignore the wrinkle recognition result of the human face image in a case where a face area corresponding to the wrinkle recognition result of the human face image is an area of the five sense organs that has no wrinkle.

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Guo Zhizhi
• Sun Yipeng
• Liu Jingtuo
• Han Junyu

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06T2207/30201

  Face · CPC title