Real-Time Object Detection Algorithm of Autonomous Vehicles Based on Improved YOLOv5s

Object detection is one of the basic task of autonomous vehicles. Object detection of high accuracy and fast detection speed are conducive to improve the safety of autonomous vehicles. In view of the large number of small objects in actual road traffic, this paper proposes a real-time object detection algorithm based on improved YOLOv5s. By adding shallow high-resolution features and changing the size of output feature map, the detection ability of the algorithm for small objects is significantly improved. The mean Average Precision of the improved YOLOv5s algorithm on BDD100K dataset increased by 3.2 percentage points, and the average detection speed is 74.6 FPS. The experimental results show that the improved YOLOv5s algorithm enhances the detection ability of small objects and proves its feasibility in various complex road scenes.