Automatic driving target detection algorithm based on improved Yolov5

The invention relates to the technical field of automatic driving, and provides an automatic driving target detection algorithm based on improved Yolov5 on the basis of deep learning. According to the algorithm, on the basis of Yolov5, a lightweight attention module SimAM is introduced into a backbone network, and weight calculation is accelerated on the premise that the parameter quantity is not increased; a PANet (path aggregation network) structure in an original Yolov5 network is replaced by a BiFPN (bidirectional feature pyramid) structure; and finally, a lightweight upsampling operator CARAFE is used, so that the whole network obtains a larger receptive field while not introducing excessive calculation amount. Therefore, the network can achieve high detection precision on the premise of ensuring the real-time performance. According to the method, the precision of targets such as vehicles and pedestrians can be improved in an automatic driving target detection task, and relatively good performance is ach