Target detection training method and system, electronic equipment and computer readable storage medium

A technology of target detection and training method, applied in the field of deep learning, can solve the problem of immaturity, and achieve the effect of improving accuracy and high detection performance

Pending Publication Date: 2021-01-22
SHANGHAI MININGLAMP ARTIFICIAL INTELLIGENCE GRP CO LTD
View PDF8 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The present invention aims at the problem that there is no mature method for semi-supervised learning in the field of target detection, and proposes a target detection training method, system, electronic equipment and computer-readable storage medium based on semi-supervised learning

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
  • Target detection training method and system, electronic equipment and computer readable storage medium
  • Target detection training method and system, electronic equipment and computer readable storage medium
  • Target detection training method and system, electronic equipment and computer readable storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0059] refer to Figure 1 to Figure 5 As shown, this example discloses a specific implementation of a semi-supervised learning-based target detection training method (hereinafter referred to as the "method").

[0060] This technical method uses the Faster RCNN network structure as the most basic training framework. Faster RCNN is a network structure with very high performance and stability in the field of target detection, such as Figure 4 As shown, this method is a two-stage network structure. First, for the image to be predicted, the candidate area is extracted, that is, the area where there may be an object of interest in the network is found, and then the extracted candidate area is extracted. For the prediction of the stage, predict its category and the exact coordinate information of its location. When there is only supervised data, the network structure uses a strategy of two calculations to obtain more accurate results. In the first stage of the RPN network, the cate...

Embodiment 2

[0092] In combination with the semi-supervised learning-based target detection training method disclosed in Embodiment 1, this embodiment discloses a specific implementation example of a semi-supervised learning-based target detection training system (hereinafter referred to as "system").

[0093] refer to Image 6 As shown, the system includes:

[0094] Feature extraction module 11: perform feature extraction on the image through the backbone to obtain image features;

[0095] One-stage noise addition module 12: the image feature obtains noise feature by noise method;

[0096] RPN network optimization module 13: optimize the RPN network using a consistent regularization method according to the image features and the noise features;

[0097] proposal output module 14: the RPN network output proposals after the image features are optimized;

[0098] Two-stage noise addition module 15: the proposals obtain noise proposals through a noise method;

[0099] roi_heads network op...

Embodiment 3

[0109] to combine Figure 7 As shown, this embodiment discloses a specific implementation manner of a computer device. The computer device may comprise a processor 81 and a memory 82 storing computer program instructions.

[0110] Specifically, the processor 81 may include a central processing unit (CPU), or an Application Specific Integrated Circuit (ASIC for short), or may be configured to implement one or more integrated circuits in the embodiments of the present application.

[0111] Among them, the memory 82 may include mass storage for data or instructions. For example without limitation, the memory 82 may include a hard disk drive (Hard Disk Drive, referred to as HDD), a floppy disk drive, a solid state drive (SolidState Drive, referred to as SSD), flash memory, optical disk, magneto-optical disk, magnetic tape or universal serial bus (Universal Serial Bus, referred to as USB) drive or a combination of two or more of the above. Storage 82 may comprise removable or no...

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 target detection training method and system, electronic equipment and a computer readable storage medium, and the method comprises the steps: extracting features from an image through a backbone, and obtaining an image feature; obtaining noise features of the image features through a noise method; optimizing the RPN network by using a consistency regularization method according to the image features and the noise features; outputting proposals by the RPN network of which the image features; obtaining noise proposals by the proposals through a noise method; optimizingthe roi_heads network by using a consistency regularization method according to the proposals and the noise proposals, and optimizing the roiheads network by using a consistency regularization methodaccording to the roiheads and the noise proposals; and obtaining a final full classification result and a final position result through the roiheads network by the proposals. On the basis of a consistency regularization technical route in the field of image classification, unlabeled data can be used for network training in the field of target detection, so that the precision of a network structureis improved, and higher detection performance is achieved.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a target detection training method, system, electronic equipment and computer-readable storage medium based on semi-supervised learning. Background technique [0002] Deep learning is currently the most commonly used and most important technical means in the field of computer vision. Deep learning uses a large amount of manually labeled image data for iterative training to achieve specific image tasks, such as image classification / target detection / semantic segmentation / image retrieval and other tasks. [0003] Among them, the target detection task is an important branch of deep learning. In this field, there are already many excellent network structures that can achieve performance comparable to human beings, such as Faster RCNN / YOLO / SSD. However, training an excellent target detector requires a large amount of manually labeled data, and the complexity of data label...

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): G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V10/44G06N3/045G06F18/241
Inventor 朱彦浩胡郡郡唐大闰
Owner SHANGHAI MININGLAMP ARTIFICIAL INTELLIGENCE GRP CO LTD
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