google/stereo-magnification

Question about training

Albert-learner opened this issue · 0 comments

Albert-learner commented

Hello, I have a question about stereo-magnification Neural Network. When I see mpi.py source code, you allocate first D(D = 32) layers to blending weights, second D(D = 32) layers to Alpha images, and last three layers for Background color image.
Could you tell me why divide layers like this order? Can I change this order? Thank you.