Michael S. (1) [Avatar] Offline
On page 122 it says:

"You can see above that the output of every Conv2D and MaxPooling2D layer is a 2D tensor of shape..."

Shouldn't it say is a 3D tensor? As it says in the text there is height, width and channels)
Vaibhav Aparimit (4) [Avatar] Offline
I think the author wanted to differentiate between kernel depth and kernel map (which is always 2D)