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

Target detection method based on SSA sharpening attention mechanism

A target detection and attention mechanism technology, applied in the field of computer vision, can solve problems such as loss of edges of large targets and difficulty in locating small and medium targets

Active Publication Date: 2021-09-14
HEFEI LONGTUTEM INFORMATION TECH CO LTD
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The present invention proposes a target detection method based on the SSA sharpening attention mechanism in order to solve the problem of difficult positioning of small and medium targets and the problem of edge loss of large targets by the lightweight target detection model

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 method based on SSA sharpening attention mechanism
  • Target detection method based on SSA sharpening attention mechanism
  • Target detection method based on SSA sharpening attention mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0034] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0035] Step (1): The input image size is 640×640; use the first-order lightweight target detection network YOLOv5s;

[0036] Step (2): Combine the SSA spatial sharpening module with the SE and ECA channel attention modules, and divide them into two types of attention modules: channel attention module and hybrid attention module, such as figure 2As shown; the channel attention module is embedded into all large residual blocks and FPN-like structures of the skeleton network of the one-stage lightweight neural network; the channel attention module is connected in series with the SSA spatial sharpening module and embedded into the one-stage lightweight neural network The nodes in the output part of the neural network, if the neural network includes a downsampling structure, are embedded before the downsampling node, here the shallow feat...

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 method based on an SSA sharpening attention mechanism, in the method, user-defined sharpening filtering is introdued into a space attention module for the first time, and combined with a channel attention module for use, thereby reducing the influence of interference factors on the SSA sharpening effect, and serving the sharpening effect; according to the invention, the edge information of the detected object in the neural network is enhanced spatially, and object positioning is enhanced. Edge information of large objects can be perfected, existence of small and medium objects in an output layer can be improved, and the detection effect is improved; according to the invention, the combination mode and the embedding position of the SSA space sharpening module and the channel attention module are perfected. Compared with a space attention module in the CBAM, the effect on the lightweight target detection model is better. According to the SSA space sharpening module, the required calculation amount and parameter amount are extremely small, the detection speed is hardly influenced, and the lightweight module is high in practicability, plug-and-play and easy to implement.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a lightweight spatial attention module based on a neural network and a sharpening attention mechanism and a target detection method combined with a channel attention module on a one-stage lightweight neural network. Background technique [0002] Document 1 (Hu J, Shen L, Sun G. Squeeze-and-excitation networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 2018.7132-7141) introduces channel attention after the network intermediate feature map The channel information of the feature map is extracted through the average pooling operation, and the global features at the channel level are obtained. Then, the connection between each channel is learned through the FC multi-layer perceptron with a hidden layer, and the weights of different channels are also obtained. After the limit of the output is limited by the Si...

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
IPC IPC(8): G06K9/00G06N3/04G06N3/08G06T5/00
CPCG06N3/08G06T2207/10004G06N3/045G06T5/73
Inventor 薛梦凡陈明皓彭冬亮杨岗贾士绅陈怡达
Owner HEFEI LONGTUTEM INFORMATION TECH CO LTD
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