A Lightweight Target Detection Algorithm for Camouflaged Soldiers
Addressing the issues of large model parameters and slow inference speed in existing models,a lightweight target detection algorithm for camouflaged soldiers are proposed.The backbone of the algorithm is designed based on the HGNetv2,SRepVGG module is utilized to fuse multi-scale features.Finally,partial convolution and 1×1 convolution are combined in the coupled detection head.The deep learning network pro-posed in the article is compared with the baseline model YOLOv8,the parameters are induced by 35.4%and inference speed is increased by 18.9%,while detection accuracy is ensured.This makes it more suitable for the operation on edge computing devices with limited computational resources.
万方数据
维普
