Hi, I am new to this. There is this problem I have encountered, I was trying to classify the pixels in the input image into classes, sort of like segmentation, using CNN. The “interested” pixels usually locate in the top right corner, but the input pictures are too big, which I have to slice them in patches, by doing this, each input patch loses its “which region of the whole picture it’s from” information. How should I manually add this info into the input, in order to let the CNN taking position into account? Thank you.
You could try to add additional channels containing the coordinate information as described in this paper from Uber.
Here is also the corresponding video explanation:
1 Like