Hey people, I’m reviewing the O’Reilly Deep Learning for Coders w/Fastai & PyTorch book and working through the CNN chapter (starting at page 403). In the book the top_edge kernell is a tensor like this
[[-1,-1,-1],
[0,0,0],
[1,1,1]]
and when applied to one of the examples of an image of a ‘3’ from MNIST_Sample the book has the output as detecting top edges, but when I run it with that kernel above I am getting an output that detects bottom edges, when I swap the '-1’s to the bottom row and the positive '1’s to the top row then it outputs top edge detection. Is this an error in the book or am I missing something?