Face detection model training method and device, electronic equipment and storage medium

A face detection and model training technology, applied in the field of image recognition, which can solve the problems of image inclusion, inconsistent picture quality, and inability to effectively identify face images.

Pending Publication Date: 2022-05-31
AGRICULTURAL BANK OF CHINA
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In the prior art, the face data set used as a training face detection model contains data images collected from different natural scenes. Due to factors such as different lighting, different imaging angles, and differences in the appearance of different people, there are usually blurred images in the image. , occluded and multi-scale dense faces, resulting in inconsistent image quality. In the case of poor sample quality, the trained face detection model cannot effectively recognize face images

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Face detection model training method and device, electronic equipment and storage medium
  • Face detection model training method and device, electronic equipment and storage medium
  • Face detection model training method and device, electronic equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0030] figure 2 The first embodiment of the present invention provides a flow chart of a face detection model training method. This embodiment can be applied to the case of recognizing facial images through a face detection model. The method can be performed by a face detection model training device. The The training device may be implemented in the form of hardware and / or software, and the hardware may be an electronic device, such as a mobile terminal or a PC.

[0031] Before introducing the solution of this embodiment, the one-stage detection algorithm in target detection can be introduced first. One-stage is an end-to-end target detection algorithm, which does not need to filter the target area in advance. Through the backbone network of the algorithm Directly complete the regression and classification of the target area. The main step of the algorithm is to complete the feature extraction of the detected object through the convolutional neural network, and then directly...

Embodiment 2

[0084] Figure 4 This is a flowchart of a method for training a face detection model provided in Embodiment 2 of the present invention, and this embodiment is a preferred embodiment of the foregoing embodiment. The specific implementation can refer to the technical solution of this embodiment. Wherein, the technical terms that are the same as or corresponding to the above embodiments are not repeated here.

[0085] like figure 2 As shown, the method includes:

[0086] (1) First choose a target detection network of the one-stage scheme as a training model, and then train two networks (as shown in Model 1 and Model 2) at the same time with the same structure and loss function. The initialization parameters are different.

[0087] (2) Independent training stage: model 1 and model 2 are trained independently for a period of time, model 1 is trained according to steps (1.1->1.2->1.4), and model 2 is trained according to steps (2.1->2.2->2.4), so that the two Each network mode...

Embodiment 3

[0093] Figure 5 This is a schematic structural diagram of a face detection model training apparatus provided in Embodiment 3 of the present invention. like Figure 5 As shown, the device includes:

[0094] The sample set obtaining module 310 is used to obtain a training sample set; wherein, the training sample set includes at least three types of training samples, and the at least three types include the correct face annotation type, the abnormal face annotation type, and the face unmarked type;

[0095] The initial model determination module 320 is configured to perform training processing on the first face detection model to be trained and the second face detection model to be trained based on the training set sample set, so as to obtain the first initial face detection model and the second initial face detection model; wherein, The model structures of the first face detection model to be trained and the second face detection model to be trained are the same, and the inde...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a face detection model training method and device, electronic equipment and a storage medium. A training sample set is obtained; wherein the training sample set comprises at least three types of training samples: a face correct labeling type, a face abnormal labeling type and a face non-labeling type; performing training processing on the first to-be-trained face detection model and the second to-be-trained face detection model based on the training set sample set to obtain a first initial face detection model and a second initial face detection model; wherein the first to-be-trained face detection model and the second to-be-trained face detection model have the same model structure, and index parameters of all neural network layers in the model structures are different; and carrying out joint training on the first initial face detection model and the second initial face detection model based on the retraining sample set to obtain a target face detection model, and carrying out detection processing based on a face image in the target face detection model image. The method achieves the accurate recognition of the image under the condition that the image quality is poor.

Description

technical field [0001] The present invention relates to the technical field of image recognition, and in particular, to a face detection model training method, device, electronic device and storage medium. Background technique [0002] In today's society, face recognition has become an identity authentication technology with the most potential and a wide range of application scenarios. As the basis of the entire face recognition system, the face detection task will have a direct impact on the final recognition result. [0003] In the prior art, the face data set used as a training face detection model contains data images collected from different natural scenes. Due to factors such as different lighting, different imaging angles, and differences in the appearance of different people, blurry images usually appear. , occluded and multi-scale dense faces, resulting in inconsistent image quality. In the case of poor sample quality, the trained face detection model cannot effect...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06V40/16G06V10/25G06V10/774G06V10/82G06K9/62
CPCG06F18/214
Inventor 李香熊辉喻潇李科
Owner AGRICULTURAL BANK OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap