R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes
A lightweight rotational object detection algorithm, R-YOLOv5, is proposed to address the limitations of traditional object detection algorithms that do not consider the diversity of vehicle scales in drone images and fail to obtain information on rotation angles. The proposed algorithm incorporated...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10083118/ |
| _version_ | 1797796522833739776 |
|---|---|
| author | Zhengwei Li Chengxin Pang Chenhang Dong Xinhua Zeng |
| author_facet | Zhengwei Li Chengxin Pang Chenhang Dong Xinhua Zeng |
| author_sort | Zhengwei Li |
| collection | DOAJ |
| description | A lightweight rotational object detection algorithm, R-YOLOv5, is proposed to address the limitations of traditional object detection algorithms that do not consider the diversity of vehicle scales in drone images and fail to obtain information on rotation angles. The proposed algorithm incorporated an angle prediction branch and introduced a circular smooth label (CSL) angle classification method to make it suitable for detection scenarios based on rotational boxes. A cascaded Swin Transformer block (STrB) is used to reduce computational complexity during feature fusion in the backbone network, further enhancing semantic information and global perception capabilities for small objects. A feature enhancement attention module (FEAM) is proposed to improve the utilization of detailed information through local feature self-supervision. An adaptive spatial feature fusion structure (ASFF) is introduced, which employs features extracted from different levels of the backbone network to perform multi-scale feature fusion. The experimental results show that the detection accuracy reaches 84.91% on the Drone-Vehicle dataset and 90.23% on the UCAS-AOD remote sensing dataset. The lightweight model has a parameter count of only 2.02 million and can achieve 82.6 FPS for high-resolution images, which is significantly better than existing lightweight models and more suitable for real-time detection of rotating vehicles in dense scenes, making it suitable for deployment on a large majority of embedded platforms. |
| first_indexed | 2024-03-13T03:34:21Z |
| format | Article |
| id | doaj.art-581f9f37a6bf40d2a11e08de8ed3a3e4 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-13T03:34:21Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-581f9f37a6bf40d2a11e08de8ed3a3e42023-06-23T23:00:50ZengIEEEIEEE Access2169-35362023-01-0111615466155910.1109/ACCESS.2023.326260110083118R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense ScenesZhengwei Li0https://orcid.org/0000-0002-9065-3915Chengxin Pang1https://orcid.org/0000-0003-2481-1433Chenhang Dong2https://orcid.org/0000-0002-8949-8815Xinhua Zeng3https://orcid.org/0000-0002-3903-0392School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai, ChinaSchool of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai, ChinaSchool of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai, ChinaA lightweight rotational object detection algorithm, R-YOLOv5, is proposed to address the limitations of traditional object detection algorithms that do not consider the diversity of vehicle scales in drone images and fail to obtain information on rotation angles. The proposed algorithm incorporated an angle prediction branch and introduced a circular smooth label (CSL) angle classification method to make it suitable for detection scenarios based on rotational boxes. A cascaded Swin Transformer block (STrB) is used to reduce computational complexity during feature fusion in the backbone network, further enhancing semantic information and global perception capabilities for small objects. A feature enhancement attention module (FEAM) is proposed to improve the utilization of detailed information through local feature self-supervision. An adaptive spatial feature fusion structure (ASFF) is introduced, which employs features extracted from different levels of the backbone network to perform multi-scale feature fusion. The experimental results show that the detection accuracy reaches 84.91% on the Drone-Vehicle dataset and 90.23% on the UCAS-AOD remote sensing dataset. The lightweight model has a parameter count of only 2.02 million and can achieve 82.6 FPS for high-resolution images, which is significantly better than existing lightweight models and more suitable for real-time detection of rotating vehicles in dense scenes, making it suitable for deployment on a large majority of embedded platforms.https://ieeexplore.ieee.org/document/10083118/UAV vehiclelightweight object detectionrotating block positioningattention network |
| spellingShingle | Zhengwei Li Chengxin Pang Chenhang Dong Xinhua Zeng R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes IEEE Access UAV vehicle lightweight object detection rotating block positioning attention network |
| title | R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes |
| title_full | R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes |
| title_fullStr | R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes |
| title_full_unstemmed | R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes |
| title_short | R-YOLOv5: A Lightweight Rotational Object Detection Algorithm for Real-Time Detection of Vehicles in Dense Scenes |
| title_sort | r yolov5 a lightweight rotational object detection algorithm for real time detection of vehicles in dense scenes |
| topic | UAV vehicle lightweight object detection rotating block positioning attention network |
| url | https://ieeexplore.ieee.org/document/10083118/ |
| work_keys_str_mv | AT zhengweili ryolov5alightweightrotationalobjectdetectionalgorithmforrealtimedetectionofvehiclesindensescenes AT chengxinpang ryolov5alightweightrotationalobjectdetectionalgorithmforrealtimedetectionofvehiclesindensescenes AT chenhangdong ryolov5alightweightrotationalobjectdetectionalgorithmforrealtimedetectionofvehiclesindensescenes AT xinhuazeng ryolov5alightweightrotationalobjectdetectionalgorithmforrealtimedetectionofvehiclesindensescenes |