An improved vehicle detection model based on YOLOv5s and attention mechanism

Vehicle detection comes as a key technology in smart traffic systems and automated driving, but existing vehicle detection models face challenges in the face of small-sized vehicles and occluded vehicles. In the present work, an improved model based on the YOLOv5s and attention mechanism is proposed to resolve the problem. In the proposed model, a spatial attention module and a channel attention module are introduced, the feature pyramid network is optimized, and the loss function is improved, which substantially increases the accuracy and efficiency of vehicle detection. Our method will provide technical support for smart traffic and automated driving, and inspire further research on vehicle detection methods.