Method and device for detecting blurriness of human face in image and computer-readable storage medium

Patent metadata

Field	Value
Publication number	US-11875599-B2
Application number	US-202117393408-A
Country	US
Kind code	B2
Filing date	Aug 4, 2021
Priority date	Aug 5, 2020
Publication date	Jan 16, 2024
Grant date	Jan 16, 2024

Abstract

A method for detecting blurriness of a human face in an image includes: performing a face detection in a target image; when a human face is detected in the target image, cropping the human face from the target image to obtain a face image and inputting the face image to a first neural network model to perform preliminary detection on a blurriness of the human face in the face image to obtain a preliminary detection result; and when the preliminary detection result meets a deep detection condition, inputting the face image to a second neural network model to perform deep detection on the blurriness of the human face in the face image to obtain a deep detection result.

First claim

What is claimed is:

1. A computer-implemented method for detecting blurriness of a human face in an image, the method comprising:
performing a face detection in a target image;
in response to detection of a human face in the target image, cropping the human face from the target image to obtain a face image and inputting the face image to a first neural network model to perform preliminary detection on a blurriness of the human face in the face image to obtain a preliminary detection result; and
in response to the preliminary detection result meeting a deep detection condition, inputting the face image to a second neural network model to perform deep detection on the blurriness of the human face in the face image to obtain a deep detection result.

2. The method according to claim 1 , wherein inputting the face image to the first neural network model to perform preliminary detection on the blurriness of the human face in the face image to obtain the preliminary detection result comprises:
inputting the face image to the first neural network model to classify the blurriness of the human face in the face image to obtain a facial blurriness classification result, wherein the facial blurriness classification result comprises a first-level blurriness, a second-level blurriness, and a third-level blurriness that are arranged in descending order of blurriness;
in response to the facial blurriness classification result being a first-level blurriness, determining that the human face in the face image is blurred;
in response to the facial blurriness classification result being a second-level blurriness, determining that the preliminary detection result meets a deep detection condition; and
in response to the facial blurriness classification result being a third-level blurriness, determining that the human face in the face image is clear.

3. The method according to claim 2 , wherein inputting the face image to the second neural network model to perform deep detection on the blurriness of the human face in the face image to obtain the deep detection result comprises:
in response to the preliminary detection result meeting the deep detection condition, inputting the face image to the second neural network model to score the blurriness of the human face;
in response to a score of the blurriness of the human face being less than the preset threshold, determining that the human face in the face image is blurred; and
in response to a score of the blurriness of the human face being greater than or equal to the preset threshold, determining that the human face in the face image is clear.

4. The method according to claim 2 , wherein inputting the face image to the first neural network model to perform preliminary detection on the blurriness of the human face in the face image to obtain the preliminary detection result comprises:
in response to detection of the human face in the target image, detecting key points of the human face in the face image;
correcting the face image based on coordinates of the key points and coordinates of key points of a standard face image to obtain a corrected face image;
inputting the corrected face image to the first neural network model to perform preliminary detection on the blurriness of the human face in the corrected face image to obtain the preliminary detection result.

5. The method according to claim 3 , further comprising, before performing the face detection in the target image,
training the second neural network model, wherein an output result of the second neural network model is the score of the blurriness of the human face.

6. The method according to claim 5 , wherein training the second neural network model comprises:
obtaining a face image set comprising a plurality of training face images;
training the second neural network model based on the face image set to obtain a probability of each of the training face images being of each of predefined N-type blur levels, wherein N≥2;
calculating a score of blurriness of a human face in each of the training face images based on the probability of each of the training face images being of each of predefined N-type blur levels; and
in response to a preset loss function converging, stopping the training of the second neural network model, wherein the loss function is configured to indicate a difference between the calculated score of blurriness of a human face in each of the training face images and a pre-flagged score of blurriness of the human face in a corresponding one of the training face images.

7. The method according to claim 6 , wherein the score of blurriness of the human face in each of the training face images is calculated according to a following formula:

E
⁡

(
O
)

=

∑
N

i

⁢

y
i

×

o
i

,

where E(O) represents the score of blurriness of the human face, y i represents the probability of each of the training face images being of a predefined i-type blur level, and O i represents a score of the i-type blur level.

8. A device for detecting blurriness of a human face in an image, the device comprising:
one or more processors;
a memory; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprise:
performing a face detection in a target image;
in response to detection of a human face in the target image, cropping the human face from the target image to obtain a face image and inputting the face image to a first neural network model to perform preliminary detection on a blurriness of the human face in the face image to obtain a preliminary detection result; and
in response to the preliminary detection result meeting a deep detection condition, inputting the face image to a second neural network model to perform deep detection on the blurriness of the human face in the face image to obtain a deep detection result.

9. The device according to claim 8 , wherein the instructions for inputting the face image to the first neural network model to perform preliminary detection on the blurriness of the human face in the face image to obtain the preliminary detection result comprise:
instructions for inputting the face image to the first neural network model to classify the blurriness of the human face in the face image to obtain a facial blurriness classification result, wherein the facial blurriness classification result comprises a first-level blurriness, a second-level blurriness, and a third-level blurriness that are arranged in descending order of blurriness;
instructions for, in response to the facial blurriness classification result being a first-level blurriness, determining that the human face in the face image is blurred;
instructions for, in response to the facial blurriness classification result being a second-level blurriness, determining that the preliminary detection result meets a deep detection condition; and
instructions for, in res

Assignees

• Ubtech Robotics Corp Ltd

Inventors

• Zeng Yusheng
• Liu Yepeng
• Cheng Jun
• Pang Jianxin
• Gu Jing

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06V40/161 Primary

  Detection; Localisation; Normalisation · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/2431

  Multiple classes · CPC title