Lightweight traffic vehicle and pedestrian target detection algorithm based on improved YOLOv7-Tiny
Aiming at the problem that the traditional traffic vehicle and pedestrian target detection algorithm is limited in the actual scene application due to the large amount of data and calculation,a lightweight traffic vehicle and pedestrian target detection algorithm based on improved YOLOv7-tiny is proposed.By designing an efficient aggregation network module based on attention mecha-nism and partial convolution,the amount of parameters and computation of the model are reduced.An adaptive multi-scale feature fusion module with jump connection is designed to improve the detection ability of the model for small targets.The regression loss function based on minimum point dis-tance bounding box is used to solve the problem that the original loss function converges slowly when the aspect ratio is the same.Model pruning is used to optimize the improved model,which not only reduces the amount of parameters and calculation,but also improves the operation efficiency of the model.The experimental results show that compared with YOLOv7-tiny,the number of parameters and calculation amount of the improved model are reduced by 67.7％and 63.3％respectively,and the accuracy is increased by 0.26％.Moreover,the model size is very small,only 4.4 MB.
万方数据
维普
