Research on foggy target detection algorithm based on improved YOLOv7-tiny
In outdoor scenes,target detection is a very important technology,which is widely used in automatic driving,moni-toring target tracking and other fields.However,the outdoor weather is changeable,especially in the foggy weather scene,due to the problems of light weakening and object edge blur,the performance of the algorithm is degraded and the detection accuracy is not high.In order to solve these problems,a modified algorithm based on YOLOv7-tiny network was proposed in the study of target de-tection algorithm in foggy weather.Firstly,the attention mechanism SimAM was introduced,and the attention module SimAM was added to the feature fusion part of the network to improve the feature extraction ability of the network for blurred images.Then the original loss function CIoU of the algorithm was replaced by wIoU to improve the positioning accuracy of the anchor frame,thereby improving the accuracy of the detection as a whole.This paper has been trained and verified on the foggy dataset ug2.Under the premise that the calculation amount of the improved algorithm remains basically unchanged,mAP0.5 is increased by 1.7percentage,and mAP0.5:0.95 is increased by 3.5 percentage,which proves the effectiveness of the improved method.
NETL
NSTL
万方数据
