Yolo_v5_5.0从debug中学习

Debug

	(1,3,256,256)	(8,3,640,640)	detect(1,3,448,640) NCHW 模型用yolo.py 里面的Model()	export的时候的model 跟inference一样
0.Focus out	(1,32,128,128)	(8,32,320,320)	(1,32,224,320)
1.Conv	(1,64,64,64)	(8,64,160,160)	(1,64,112,160)
2.C3	(1,64,64,64)	(8,64,160,160)	(1,64,112,160)
3.Conv	(1,128,32,32)	(8,128,80,80)	(1,128,56,80)
4.C3	(1,128,32,32)	(8,128,80,80)	(1,128,56,80)
5.Conv	(1,256,16,16)	(8,256,40,40)	(1,256,28,40)
6.C3	(1,256,16,16)	(8,256,40,40)	(1,256,28,40)
7.Conv	(1,512,8,8)	(8,512,20,20)