An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning
An improved algorithm has been proposed to address the challenges encountered in object detection using visible and thermal infrared images. These challenges include the diversity of object detection perspectives, deformation of the object, occlusion, illumination, and detection of small objects. Th...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2023-05-01
|
Series: | Frontiers in Physics |
Subjects: | |
Online Access: | https://www.frontiersin.org/articles/10.3389/fphy.2023.1193245/full |
_version_ | 1797828842345201664 |
---|---|
author | Xiaoguang Tu Xiaoguang Tu Zihao Yuan Bokai Liu Jianhua Liu Yan Hu Houqiang Hua Lin Wei |
author_facet | Xiaoguang Tu Xiaoguang Tu Zihao Yuan Bokai Liu Jianhua Liu Yan Hu Houqiang Hua Lin Wei |
author_sort | Xiaoguang Tu |
collection | DOAJ |
description | An improved algorithm has been proposed to address the challenges encountered in object detection using visible and thermal infrared images. These challenges include the diversity of object detection perspectives, deformation of the object, occlusion, illumination, and detection of small objects. The proposed algorithm introduces the concept of contrastive learning into the YOLOv5 object detection network. To extract image features for contrastive loss calculation, object and background image regions are randomly cropped from image samples. The contrastive loss is then integrated into the YOLOv5 network, and the combined loss function of both object detection and contrastive learning is used to optimize the network parameters. By utilizing the strategy of contrastive learning, the distinction between the background and the object in the feature space is improved, leading to enhanced object detection performance of the YOLOv5 network. The proposed algorithm has shown pleasing detection results in both visible and thermal infrared images. |
first_indexed | 2024-04-09T13:10:52Z |
format | Article |
id | doaj.art-0d07171bd274463dadce069ecab7faf1 |
institution | Directory Open Access Journal |
issn | 2296-424X |
language | English |
last_indexed | 2024-04-09T13:10:52Z |
publishDate | 2023-05-01 |
publisher | Frontiers Media S.A. |
record_format | Article |
series | Frontiers in Physics |
spelling | doaj.art-0d07171bd274463dadce069ecab7faf12023-05-12T07:01:55ZengFrontiers Media S.A.Frontiers in Physics2296-424X2023-05-011110.3389/fphy.2023.11932451193245An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learningXiaoguang Tu0Xiaoguang Tu1Zihao Yuan2Bokai Liu3Jianhua Liu4Yan Hu5Houqiang Hua6Lin Wei7Institute of Electronic and Electrical Engineering, Civil Aviation Flight University of China, Guanghan, ChinaSchool of Computer Science, Sichuan University, Chengdu, ChinaInstitute of Electronic and Electrical Engineering, Civil Aviation Flight University of China, Guanghan, ChinaCollege of Aviation Engineering, Civil Aviation Flight University of China, Guanghan, ChinaInstitute of Electronic and Electrical Engineering, Civil Aviation Flight University of China, Guanghan, ChinaInstitute of Electronic and Electrical Engineering, Civil Aviation Flight University of China, Guanghan, ChinaInstitute of Electronic and Electrical Engineering, Civil Aviation Flight University of China, Guanghan, ChinaCollege of Flight Technology, Civil Aviation Flight University of China, Guanghan, ChinaAn improved algorithm has been proposed to address the challenges encountered in object detection using visible and thermal infrared images. These challenges include the diversity of object detection perspectives, deformation of the object, occlusion, illumination, and detection of small objects. The proposed algorithm introduces the concept of contrastive learning into the YOLOv5 object detection network. To extract image features for contrastive loss calculation, object and background image regions are randomly cropped from image samples. The contrastive loss is then integrated into the YOLOv5 network, and the combined loss function of both object detection and contrastive learning is used to optimize the network parameters. By utilizing the strategy of contrastive learning, the distinction between the background and the object in the feature space is improved, leading to enhanced object detection performance of the YOLOv5 network. The proposed algorithm has shown pleasing detection results in both visible and thermal infrared images.https://www.frontiersin.org/articles/10.3389/fphy.2023.1193245/fulldeep learningYOLOv5object detectioncontrastive learninginfrared thermal image |
spellingShingle | Xiaoguang Tu Xiaoguang Tu Zihao Yuan Bokai Liu Jianhua Liu Yan Hu Houqiang Hua Lin Wei An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning Frontiers in Physics deep learning YOLOv5 object detection contrastive learning infrared thermal image |
title | An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning |
title_full | An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning |
title_fullStr | An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning |
title_full_unstemmed | An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning |
title_short | An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning |
title_sort | improved yolov5 for object detection in visible and thermal infrared images based on contrastive learning |
topic | deep learning YOLOv5 object detection contrastive learning infrared thermal image |
url | https://www.frontiersin.org/articles/10.3389/fphy.2023.1193245/full |
work_keys_str_mv | AT xiaoguangtu animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT xiaoguangtu animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT zihaoyuan animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT bokailiu animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT jianhualiu animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT yanhu animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT houqianghua animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT linwei animprovedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT xiaoguangtu improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT xiaoguangtu improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT zihaoyuan improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT bokailiu improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT jianhualiu improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT yanhu improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT houqianghua improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning AT linwei improvedyolov5forobjectdetectioninvisibleandthermalinfraredimagesbasedoncontrastivelearning |