Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Target detection network and method based on mixed cavity convolution pyramid

A technology of target detection and pyramid, which is applied in the field of target detection network based on hybrid hole convolution pyramid, can solve the problems of affecting detection results, redundant feature map fusion methods, and poor backbone network performance.

Active Publication Date: 2021-09-14
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF14 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This method uses feature maps of different sizes, but the feature map fusion method is relatively redundant, and the performance of the backbone network is not good, which affects the final detection results

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 network and method based on mixed cavity convolution pyramid
  • Target detection network and method based on mixed cavity convolution pyramid
  • Target detection network and method based on mixed cavity convolution pyramid

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0051] As the basic implementation of the present invention, the present invention includes a target detection network based on a hybrid dilated convolutional pyramid, including a backbone network, a mixed receptive field module, a low-level embedded feature pyramid module, and a detection module. The backbone network uses a hierarchical and cascaded network structure to extract target picture features; the mixed receptive field module is used to enhance the features of the highest-level feature map output from the top of the backbone network. The low-level embedded feature pyramid module is used to fuse high-level features downward on the basis of the feature pyramid, and generate a final feature map to be detected by means of low-level embedding. The detection module is used to locate and classify the feature map to be detected, and output the result.

[0052] The backbone network can be a single-stage detection network based on the Res2Net50 network, which has a stronger fe...

Embodiment 2

[0059] As the best implementation mode of the present invention, the present invention includes a target detection network based on a hybrid dilated convolutional pyramid, referring to the appended figure 1 , including a backbone network, a mixed receptive field module, a low-level embedded feature pyramid module, and a detection module.

[0060] The backbone network adopts a single-stage detection network structure, introduces the Achor-free mechanism of FCOS, performs pixel-by-pixel prediction, does not rely on pre-defined anchor frames or proposed regions for target detection, and reduces the invalidity caused by redundant candidate frames Calculation improves the positioning accuracy and effectively solves problems such as missed detection. It uses its Centerness mechanism to quickly filter negative samples, suppress low-quality prediction frames far from the target center, and increase the weight of the prediction frame close to the target center. Detection performance. ...

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 relates to the technical field of digital image processing, in particular to a target detection network and method based on a hybrid cavity convolution pyramid, and the target detection network comprises a backbone network, a hybrid receptive field module, a low-layer embedded feature pyramid module and a detection module. The backbone network uses a hierarchical cascade network structure to extract target picture features; the hybrid receptive field module performs feature enhancement on the highest layer feature map output by the topmost end of the backbone network; the low-layer embedded feature pyramid module is used for downwards fusing high-layer features on the basis of a feature pyramid and generating a final to-be-detected feature map in a low-layer embedding manner; and the detection module is used for positioning and classifying the to-be-detected feature map and outputting a result. Through the target detection network and the target detection method, the problems of missing detection and error detection caused by scales and shielding can be effectively solved.

Description

technical field [0001] The present invention relates to the technical field of digital image processing, in particular to a target detection network and method based on a mixed hole convolution pyramid. Background technique [0002] Object detection is one of the most widespread applications in real life, where the task is to focus on a specific object in a picture. The traditional target detection method can be divided into a single-stage target detection method and a two-stage target detection method. The core of the two-stage method is to use the region proposal method to selectively search the input image and generate a region suggestion box, and then propose a region for each region. The box uses a convolutional neural network to extract features, and then uses a classifier for classification. The single-stage method directly outputs the target detection results through the convolutional neural network. [0003] After a series of variants, the common point of these tw...

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): G06N3/04G06N3/08G06K9/62
CPCG06N3/08G06N3/045G06F18/253
Inventor 殷光强殷康宁候少麒梁杰丁一寅曾宇昊
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com