With the rapid development of artificial intelligence technology in recent years, object detection methods have become a research hotspot in theory and application. However, the existing detection methods generally have the problem of low detection accuracy. To solve this problem, some scholars have proposed deep learning-based models, but this increases the complexity of the model and reduces the training efficiency. To this end, this paper proposes a new improved YOLOv5s algorithm that balances lightweight and performance. First, replace the original C2F module with MobileNetV3-Small to reduce the model complexity. Then, the SE attention mechanism is introduced to obtain global information, learn the correlation between features at different scales and fuse them, enhance the semantic information of features, and use SGD as an optimizer to further improve the accuracy. This paper is verified on the STL-10 public data set. The experimental results show that after the introduction of the MobileNetV3-Small framework, the number of valid parameters of the model is reduced, and the training time is greatly reduced. At the same time, compared with other mechanism attention, the SE attention mechanism has the greatest improvement in performance, and has excellent performance in lightweight and algorithm performance balance. The effectiveness of the optimization strategy has been verified. Compared with the underlying Yolov5 algorithm, the proposed improved Yolov5s algorithm improves the detection accuracy by 0.5, and the superiority of the model is verified.

Chenxi Yan, Jiafeng Li. An improved yolov5s algorithm and its application in object detection. Academic Journal of Computing & Information Science (2024), Vol. 7, Issue 12: 36-43. https://doi.org/10.25236/AJCIS.2024.071205.

[1] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016.

[2] Redmon J, Farhadi A. YOLO9000: Better, faster, stronger[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, 2017.

[3] Redmon J, Farhadi A. YOLOv3: An incremental improvement[J]. Computer Vision and Pattern Recognition, 2018, 3: 121-126.

[4] Bochkovskiy A, Wang CY, Liao HY. YOLOv4: Optimal speed and accuracy of object detection[J]. Cornell Universit, 2020, 3(8): 11-16.

[5] Liu S, Zhou X, Wang Y. Research on ship recognition algorithm based on improved YOLOv5[J]. Information Technology and Informatization, 2023, (08): 188-193+198.

[6] Robbins H, Monro S. A stochastic approximation method[J]. Annals of Mathematical Statistics, 1951, 22(3): 400-407. 

[7] Li Z, Zhao Y. Face detection with improved Adam optimization algorithm[J]. Journal of Taiyuan Normal University (Natural Science Edition), 2022, 21(04): 58-63.

[8] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]. 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 2014.

[9] Carion N, Massa F, Synnaeve G, et al. End-to-end object detection with transformers[C]. Computer Vision-ECCV 2020, Lecture Notes in Computer Science, 2020: 213-229.

[10] Qin X, Zhang Z, Huang C, et al. U2-Net: Going deeper with nested U-structure for salient object detection [J]. Pattern Recognition, 2020: 107404.

[11] Han X. Quantum multi-class classification support vector machine optimized for stochastic gradient descent [J]. Fujian Computer, 2024, (4): 1-6. 

[12] Zeng Q L, Zhou G Y, Wan L R, et al. Detection of coal and gangue based on improved YOLOv8[J]. Sensors, 2024, 24(4): 1246. 

[13] Howard A, Sandler M, Chu G, et al. Searching for MobileNetV3[J]. Computer Vision and Pattern Recognition, 2019, 45(6): 589-594. 

[14] Lin S, Liu M, Tao Z. Underwater treasure detection using attention mechanism and improved YOLOv5[J]. Chinese Journal of Agricultural Engineering, 2021, (18): 307-314.