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

Patent metadata

Field	Value
Publication number	US-11756332-B2
Application number	US-202117208568-A
Country	US
Kind code	B2
Filing date	Mar 22, 2021
Priority date	Jun 30, 2020
Publication date	Sep 12, 2023
Grant date	Sep 12, 2023

Abstract

The present application discloses an image recognition method, apparatus, device, and a computer storage medium, which is related to a technical field of artificial intelligence, and in particular, to a technical field of image processing. The method includes: performing organ recognition on a human face image and marking positions of the human facial five sense organs in the human face image, obtaining a marked human face image; inputting the marked human face image into a backbone network model and performing feature extraction, obtaining defect features of the marked human face image outputted by different convolutional neural network levels of the backbone network model; and fusing the defect features of different levels that are located in a same area of the human face image, obtaining a defect recognition result of the human face image.

First claim

What is claimed is:

1. An image recognition method, comprising:
performing organ recognition on a human face image and marking positions of human facial five sense organs in the human face image, obtaining a marked human face image;
inputting the marked human face image into a backbone network model and performing feature extraction, obtaining defect features of the marked human face image outputted by different convolutional neural network levels of the backbone network model; and
fusing the defect features of different levels that are located in a same area of the human face image, obtaining a defect recognition result of the human face image,
wherein the inputting the marked human face image into a backbone network model and performing feature extraction, obtaining defect features of the marked human face image outputted by different convolutional neural network levels of the backbone network model, comprises:
setting a priori box on the marked human face image in the convolutional neural network level of a target level, wherein the target level is one of the multiple levels in the convolutional neural network in the backbone network model, and the size of the priori box corresponds to the target level; and
determining whether there are human facial defects in the priori box, and outputting a partial human face image in the priori box as a defect feature of the marked human face image outputted by the convolutional neural network level of the target level if it is determined that there are human facial defects in the priori box.

2. The method of claim 1 , wherein prior to the performing organ recognition on a human face image, the method further comprises:
performing homogenization processing on pixels of the human face image, obtaining a homogenized human face image; and
performing normalization processing on pixel variances of the homogenized human face image, obtaining a normalized human face image.

3. The method of claim 1 , wherein the determining whether there are human facial defects in the priori box comprises:
determining that the partial human face image in the priori box does not have human facial defects in a case that a human face part corresponding to the partial human face image in the priori box is a target part, wherein the target part is a part of the marked human facial five sense organs that does not have defects; and
obtaining an image feature of the partial human face image in the priori box and then determining whether there are human facial defects based on the image feature of the partial human face image in the priori box in a case that the human face part corresponding to the partial human face image in the priori box is not the target part.

4. The method of claim 3 , wherein the determining whether there are human facial defects based on the partial human face image in the priori box further comprises:
moving the priori box by a set offset on the basis of the current position in a case that it is determined that the partial human face image in the priori box does not have human facial defects and re-executing the step of determining whether there are human facial defects based on the partial human face image in the priori box.

5. An image recognition apparatus, comprising:
a processor and a memory for storing one or more computer programs executable by the processor,
wherein when executing at least one of the computer programs, the processor is configured to perform operations comprising:
performing organ recognition on a human face image and mark positions of the human facial five sense organs in the human face image, obtaining a marked human face image;
inputting the marked human face image into a backbone network model and perform feature extraction, obtaining defect features of the marked human face image outputted by different convolutional neural network levels of the backbone network model; and
fusing the defect features of different levels that are located in a same area of the human face image, obtaining a defect recognition result of the human face image,
wherein when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
setting the priori box on the marked human face image in the convolutional neural network level of a target level, wherein the target level is one of the multiple levels in the convolutional neural network in the backbone network model, and the size of the priori box corresponds to the target level; and
determining whether there are human facial defects in the priori box and output a partial human face image in the priori box as a defect feature of the marked human face image outputted by the convolutional neural network level of the target level if it is determined that there are human facial defects in the priori box.

6. The apparatus of claim 5 , wherein when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
performing homogenization processing on pixels of the human face image, obtaining a homogenized human face image; and
performing normalization processing on pixel variances of the homogenized human face image, obtaining a normalized human face image.

7. The apparatus of claim 5 , wherein when executing at least one of the computer programs, the processor is further configured to perform operations comprising:
determining that the partial human face image in the priori box does not have human facial defects in a case that a human face part corresponding to the partial human face image in the priori box is a target part, wherein the target part is a part of the marked human facial five sense organs that does not have defects; and
obtaining an image feature of the partial human face image in the priori box and then determine whether there are human facial defects based on the image feature of the partial human face image in the priori box in a case that the human face part corresponding to the partial human face image in the priori box is not the target part.

8. The apparatus of claim 7 , wherein when executing at least one of the computer programs, the processor is further configured to perform operations comprising: moving the priori box by a set offset on the basis of the current position in a case that it is determined that the partial human face image in the priori box does not have human facial defects and re-execute the step of determining whether there are human facial defects based on the partial human face image in the priori box.

9. A non-transitory computer-readable storage medium having computer instructions stored therein, wherein the computer instructions, when executed by a computer, cause the computer to perform the method according to claim 1 .

10. A non-transitory computer-readable storage medium having computer instructions stored therein, wherein the computer instructions, when executed by a computer, cause the computer to perform the method according to claim 2 .

11. A non-transitory computer-readable storage medium having computer instructions stored therein, wherein the computer instructions, when executed by a computer, cause the computer to perform the method according to claim 3 .

12. A non-transitory computer-readable storage medium having computer instructions stored therein, wherein the computer instructions, when executed by a computer, cause the computer to perform the method according to claim 4 .

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

• G06V10/82

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06V40/171 Primary

  Local features and components; Facial parts (eye characteristics G06V40/18); Occluding parts, e.g. glasses; Geometrical relationships · CPC title