23 经典卷积神经网络 LeNet

网络架构图

Layer	Description	Input Size	Output Size	Weight	Flop
Input	32×32 grayscale image	1×1×32×32	1×1×32×32	0	0
C1：Conv	Conv layer with 6filters of size 5×5，stride 1	1×1×32×32	1×6×28×28	1×6×5×5+6	1×5×5×outputsize
S2：Pooling	Pooling layer with 2×2 average pooling，stride 2	1×6×28×28	1×6×14×14	0	0
C3：Conv	Conv layer with 16 filters of sizen 5×5，stride1	1×6×14×14	1×16×10×10	6×16×5×5+16	6×5×5×outputsize
S4:Conv	Conv layer with 2×2 average pooling，stride 2	1×16×10×10	1×16×5×5	0	0
F5:Fully connected layer	Fully connected layer with 400 inputs,120 outputs	400	120	400×120+120	400×120
F6：Fully connected layer	Fully connected layer with 120 inputs，84 outputs	120	84	120×84+84	120×84
output	output layer with 84 inputs，10 outputs（10-class）	84	10	84×10+10	84×10