Research Article |
YOLOv8-SOR: A Small-Object-Responsive Road Obstacle Detection Model using RepVGG and Swin Transformer
Author(s): LiKang Bo1,2*, Fei Lu Siaw1, Tzer Hwai Gilbert Thio1
Published In : International Journal of Electrical and Electronics Research (IJEER) Volume 13, Issue 2
Publisher : FOREX Publication
Published : 30 August 2025
e-ISSN : 2347-470X
Page(s) : 419-428
Abstract
Accurate and efficient road obstacle detection is crucial for ensuring the safety and reliability of autonomous and assisted driving systems. However, current object detection models, including YOLOv8, often encounter difficulties in accurately detecting small obstacles and balancing detection precision with inference speed. To address these challenges, we propose an improved YOLOv8-based detection framework featuring three key enhancements: (1) the backbone is redesigned using RepVGG modules to accelerate inference through structural re-parameterization without compromising feature representation; (2) an additional small-object detection head at the P2 feature level is introduced to significantly enhance sensitivity towards small-scale obstacles; (3) a Swin Transformer module is integrated into the final stage of the backbone to improve global contextual understanding and semantic feature extraction. Extensive experiments conducted on a comprehensive, multi-source road obstacle dataset demonstrate that our proposed model achieves superior performance with a precision of 0.812, recall of 0.782, mAP@0.5 of 0.822, and mAP@0.5:0.95 of 0.588. Additionally, it maintains a real-time inference speed of 180 FPS on an NVIDIA H800 GPU. The results confirm the efficacy of our approach, particularly in enhancing small object detection in complex real-world traffic scenarios.
Keywords: Road Obstacle Detection
, Small Object Detection
, RepVGG
, Transformer
.
LiKang Bo, Centre for Sustainability in Advanced Electrical and Electronic Systems (CSAEES), Faculty of Engineering, Built Environment and Information Technology, SEGi University, 47810 Petaling Jaya, Selangor, Malaysia;
Fei Lu Siaw , Centre for Sustainability in Advanced Electrical and Electronic Systems (CSAEES), Faculty of Engineering, Built Environment and Information Technology, SEGi University, 47810 Petaling Jaya, Selangor, Malaysia;
Tzer Hwai Gilbert Thio, Centre for Sustainability in Advanced Electrical and Electronic Systems (CSAEES), Faculty of Engineering, Built Environment and Information Technology, SEGi University, 47810 Petaling Jaya, Selangor, Malaysia;
LiKang Bo,Hebei Vocational University of Technology and Engineeringļ¼No.473, Quannan West Street, Xindu District, Xingtai, Hebei, China
-
[1] Wang, C., B. Zheng, and C. Li, Efficient traffic sign recognition using YOLO for intelligent transport systems. Scientific Reports, 2025. 15(1).
-
[2] Juyal, A., S. Sharma, and S. Bhadula, Modified-vehicle detection and localization model for autonomous vehicle traffic system. Indonesian Journal of Electrical Engineering and Computer Science, 2025. 37(2): p. 1183-1200.
-
[3] Redmon, J., et al. You only look once: Unified, real-time object detection. in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2016.
-
[4] Ding, X., et al. RepVgg: Making VGG-style ConvNets Great Again. in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2021.
-
[5] Liu, Z., et al. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. in Proceedings of the IEEE International Conference on Computer Vision. 2021.
-
[6] Liu, G., et al., Lightweight obstacle detection for unmanned mining trucks in open-pit mines. Scientific Reports, 2025. 15(1).
-
[7] Choi, E., T.A. Dinh, and M. Choi, Enhancing Driving Safety of Personal Mobility Vehicles Using On-Board Technologies. Applied Sciences (Switzerland), 2025. 15(3).
-
[8] Zhou, S., et al. Small Target Detector Based on Adaptive Re-parameterized Spatial Feature Fusion Mechanism. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2025.
-
[9] Zhou, X., Chen, W., & Wei, X. (2024). Improved Field Obstacle Detection Algorithm Based on YOLOv8. Agriculture, 14(12), 2263. https://doi.org/10.3390/agriculture14122263.
-
[10] Y. Liu, H.Z., H. Liu and M. Zhao, "HSC-YOLOv7: An Enhanced YOLOv7 for Small Object Detection," 2024 43rd Chinese Control Conference (CCC), Kunming, China, 2024, pp. 8910-8915, doi: 10.23919/CCC63176.2024.10661994.
-
[11] Ma, S., Zhao, X., Wan, L. et al. A lightweight algorithm for steel surface defect detection using improved YOLOv8. Sci Rep 15, 8966 (2025). https://doi.org/10.1038/s41598-025-93469-5.
-
[12] Chen, J., Mai, H., Luo, L., Chen, X., & Wu, K. (2021, September). Effective feature fusion network in BIFPN for small object detection. In 2021 IEEE international conference on image processing (ICIP) (pp. 699-703). IEEE.
-
[13] Gong, Y., Yu, X., Ding, Y., Peng, X., Zhao, J., & Han, Z. (2021). Effective fusion factor in FPN for tiny object detection. In Proceedings of the IEEE/CVF winter conference on applications of computer vision (pp. 1160-1168).
-
[14] Wu, J., & Liao, S. (2022). Traffic sign detection based on SSD combined with receptive field module and path aggregation network. Computational intelligence and neuroscience, 2022(1), 4285436.
-
[15] Woo, S., Park, J., Lee, J. Y., & Kweon, I. S. (2018). Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision (ECCV) (pp. 3-19).
-
[16] Lei, Y., Pan, D., Feng, Z., & Qian, J. (2023). Lightweight YOLOv5s human Ear recognition based on MobileNetV3 and ghostnet. Applied Sciences, 13(11), 6667.
-
[17] Hong, J., K. Ye, and S. Qiu, Study on lightweight strategies for L-YOLO algorithm in road object detection. Scientific Reports, 2025. 15(1).
-
[18] Leng, X., et al., An Improved YOLOv8-Based Method for Real-Time Detection of Harmful Tea Leaves in Complex Backgrounds. Phyton-International Journal of Experimental Botany, 2024. 93(11): p. 2963-2981.
-
[19] Alahdal, N.M., et al. Real-time Object Detection in Autonomous Vehicles with YOLO. in Procedia Computer Science. 2024.
-
[20] Zhou, X., W. Chen, and X. Wei, Improved Field Obstacle Detection Algorithm Based on YOLOv8. Agriculture (Switzerland), 2024. 14(12).
-
[21] Zhang, H., M. Liang, and Y. Wang, YOLO-BS: a traffic sign detection algorithm based on YOLOv8. Scientific Reports, 2025. 15(1).
-
[22] Jhala, J.S., C. Joshi, and D. Anand. Deep Learning Driven Object Detection and Classification for Autonomous Vehicles in Diverse Traffic and Weather Conditions. in 1st International Conference on Emerging Technologies for Dependable Internet of Things, ICETI 2024. 2024.