Abstract: Aiming at the problem of missing detection of small target vehicles in autonomous driving, this paper proposes an improved vehicle detection algorithm based on YOLOv5s. The algorithm adopts the weighted bidirectional feature pyramid network (BiFPN) fusion method, which can enhance the fusion of different levels of information while preserving more shallow semantic information. It also introduces multiple self-attention mechanisms into the backbone network to improve the feature extraction capability. The experimental results show that, compared with the unimproved YOLOv5s model, the mean average precision (mAP) of the improved network model increases by 1.01%. Its detection speed meets the real-time requirements, and it can effectively detect small target vehicles under different lighting conditions.