Research on Automatic Driving Target Detection Based on YOLOv5s. Issue 1 (1st January 2022)

• Record Type: Journal Article
• Authors: Li, Chao; Cao, Yining; Peng, Yakun
• Is Part Of: Journal of physics. Volume 2171:Issue 1(2022)
• Publication Date: 2022-01-01
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/2171/1/012047
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Aiming at the problems of low detection accuracy, small targets and inaccurate recognition of automatic driving in natural scenes such as different lighting, different weather and different perspectives. The yolov5s network, migration learning method and varifocal loss function are used, and the hyper-parameter adjustment and network fine-tuning are completed. Experimental results show that the proposed method achieves mean Average Precision of 46.68%, and the detection speed can reach 7ms/frame. Automatic driving target detection based on YOLOv5s model network has fast training results, easy deployment, strong robustness, and high recognition accuracy in different weather conditions. It can not only meet the accuracy of automatic driving target detection, but also meet the real-time detection.