How does the convolution kernel change the number of channels?
1: If it is a 240*240 color image, it has three RGB channels.
2: Use a convolution kernel of any size to convolve with its three channels respectively. If it is a 3*3 convolution kernel, it means that the 3*3 convolution kernel is convolved with each of the three channels once, and the convolution is obtained. The three channels of , add the corresponding positions of the three channels to get a picture, which is a channel.
3: Loop this process 10 times to get 10 channels, loop it 50 times to get 50 channels.
Therefore, there are as many channels as there are convolution kernels. Each convolution kernel will be convolved with the original channel and then added to obtain a channel. How many convolution kernels are there, how many times will it be cycled?