output_size=( input_size + pad * 2 - conv_size ) / stride + 1
输入:N0*C0*H0*W0
输出:N1*C1*H1*W1
输出的feature map大小:
H1=H0+2×pad−kernel_sizestride+1
W1=W0+2×pad−kernel_sizestride+1
注:当stride为1时,若pad=kernel_size−12,那么经过计算后的feature map大小不变
output_size=( input_size + pad * 2 - conv_size ) / stride + 1
输入:N0*C0*H0*W0
输出:N1*C1*H1*W1
输出的feature map大小:
H1=H0+2×pad−kernel_sizestride+1
W1=W0+2×pad−kernel_sizestride+1
注:当stride为1时,若pad=kernel_size−12,那么经过计算后的feature map大小不变