Search

Research on Lightweight YOLO-X Vehicle Detection Model with Multi-Scale Feature Fusion and Dynamic Sample Weighting

Author(s)

Shuyang Li*

Affiliation(s)

Department of Telecommunications Engineering, Xidian University, Xi'an, Shaanxi, China *Corresponding author.

Abstract

This paper proposes a lightweight YOLO-X vehicle detection model. By introducing a multi-scale feature fusion module and a dynamic sample weighting mechanism, the detection accuracy and computational efficiency are significantly improved. The multi-scale feature fusion module effectively integrates features of different scales through cross-layer connections and adaptive weighting strategies, enhancing the adaptability to multi-scale targets; the dynamic sample weighting mechanism adaptively adjusts the sample loss weights and optimizes the learning ability of difficult samples. Experimental results show that the model achieves high detection performance (mAP of 0.86, F1 score of 0.85) while maintaining low computational complexity (FLOPs of 8.2G) and parameter volume (4.5M), and the frame rate reaches 102 FPS, meeting the needs of real-time detection. Compared with mainstream models, lightweight YOLO-X performs well in both accuracy and efficiency, providing an efficient solution for vehicle detection in resource-constrained environments, and is suitable for fields such as intelligent transportation and autonomous driving.

Keywords

Lightweight YOLO-X; Multi-Scale Feature Fusion; Dynamic Sample Weighting; Vehicle Detection; Real-Time Performance

References

[1] Qian, X., Wang, X., Yang, S., Lei, J.: LFF-YOLO: A YOLO Algorithm With Lightweight Feature Fusion Network for Multi-Scale Defect Detection. IEEE Access 10, 130339-130349 (2022). [2] Shen, X., Li, H., Huang, Y., Wang, Y.: Vehicle detection method based on adaptive multi-scale feature fusion network. Journal of Electronic Imaging 31, 043008-043008 (2022). [3] Zhang, H., Li, G., Wan, D., Wang, Z., Dong, J., Lin, S., Deng, L., Liu, H.: DS-YOLO: A dense small object detection algorithm based on inverted bottleneck and multi-scale fusion network. Biomimetic Intelligence and Robotics (2024). [4] Feng, J., Wang, J., Qin, R.: Lightweight detection network for arbitrary-oriented vehicles in UAV imagery via precise positional information encoding and bidirectional feature fusion. International Journal of Remote Sensing 44, 4529-4558 (2023). [5] Tao, W., Wang, X., Yan, T., Liu, Z., Wan, S.: ESF-YOLO: an accurate and universal object detector based on neural networks. Frontiers in Neuroscience 18 (2024). [6] Zhao, X., Chen, Y.: YOLO-DroneMS: Multi-Scale Object Detection Network for Unmanned Aerial Vehicle (UAV) Images. Drones (2024). [7] Luo, Q., Wang, J., Gao, M., Lin, H., Zhou, H., Miao, Q.: G-YOLOX: A Lightweight Network for Detecting Vehicle Types. J. Sensors 2022, 1-10 (2022). [8] Peng, G., Yang, Z., Wang, S., Zhou, Y.: AMFLW-YOLO: A Lightweight Network for Remote Sensing Image Detection Based on Attention Mechanism and Multiscale Feature Fusion. IEEE Transactions on Geoscience and Remote Sensing 61, 1-16 (2023). [9] Guozheng Nan, Yue Zhao, Chengxing Lin, Qiaolin Ye: General Optimization Methods for YOLO Series Object Detection in Remote Sensing Images. IEEE Signal Processing Letters 31, 2860-2864 (2024). [10] Tianyi Xie, Wen-bo Han, Sheng Xu: YOLO-RS: A More Accurate and Faster Object Detection Method for Remote Sensing Images. Remote Sensing (2023). [11] Qi Zhang, Hongying Zhang, Xiu Lu: Adaptive Feature Fusion for Small Object Detection. Applied Sciences (2022).