Hello,
I’m trying to implement a region proposal network and need a sliding convolutional layer, but would like to adjust it to focus on the diagonal portions of the network. Is having a stride of (1,1) in a tuple the equivalent of just making predictions along the diagonal?