Research on Infrared Small Target Detection Algorithm and Model Lightweight

Shuo Han*, Zhiqiang Guo, Bo Mo, Jie Zhao

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

In order to improve the detection accuracy of the existing algorithm YOLOv5s for small infrared targets and lightweight its model, YOLOv5s-SIT algorithm is designed to optimize the network structure, including ECBAM attention module that enhances important features and suppresses non-important features, and spp_x fusion module to enrich the expression ability of feature maps etc. The lightweight model first performs sparse training on the BN layer, and then prunes and compresses the feature extraction backbone network of the algorithm without changing the integrity of the model. Experimental results show that the algorithm can achieve higher detection accuracy, faster detection speed, and lightweight model volume.

源语言英语
主期刊名Proceedings - 2023 38th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2023
出版商Institute of Electrical and Electronics Engineers Inc.
532-538
页数7
ISBN(电子版)9798350303636
DOI
出版状态已出版 - 2023
活动38th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2023 - Hefei, 中国
期限: 27 8月 202329 8月 2023

出版系列

姓名Proceedings - 2023 38th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2023

会议

会议38th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2023
国家/地区中国
Hefei
时期27/08/2329/08/23

指纹

探究 'Research on Infrared Small Target Detection Algorithm and Model Lightweight' 的科研主题。它们共同构成独一无二的指纹。

引用此