id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
fcis-6024	He, Xing	Vehicle target detection algorithm based on yolov5	2023	4	.pdf	application/pdf	3007	126	49	al.[6]borrowed the practice of global feature extraction of feature maps in SPPNet, and improved on the basis of R-CNN to propose Fast R-CNN, which solved the problem of cumbersome training of R-CNN and each training stage being independent of each other, and secondly solved the problem of requiring consistent input image size, and finally replaced the original SVM classifier with softmax. ACmix Hybrid Self-Attention Structure 58 Traditional convolution and self-attention modules actually share the same operation when projecting input feature maps by 1×1 convolution, and these two operations take up the main computational overhead.	cache/fcis-6024.pdf	txt/fcis-6024.txt