Convolution Neural Network

For example, we are taking a 3030 RGB image. So the image shape will be 3030*3.

For the first convolution layer, we are taking 10 filters. And will keep padding in a way the image size is preserved.

So now the size will be 303010

Now, for the second convolution layer, we are taking 5 filters. I want to understand how this convolution operation will take place. Like each filter will go to all 10 filters of layer one?

Hi @adit_doshi,
Yes, though there are other filters as well.
The one you’re referring to is called a 2D conv with 3D input.
Here, each of the 5 filters will take in n*n*10 as an input, and give us 1*1 as output in one pass.
The reason it is called a 2D conv is the fact that it is allowed to move in only two directions.