Convolution basics

Convolution Kernels

• 1 * 1 convolution layer is used to change the number of channel

• To transform a feature map from $C_1$ channels to $C_2$ channels, we will need $C_1 * C_2$ kernels

• Stride, padding, and kernel size determine the width and height of the output feature map :

• Pooling changes the size of feature map but not the channel number of it.