about the implementation..

[chenchr]

Hello. Thanks for the paper and code. I have done some work about bilateral filter before and I am curious about the torch implementation of dpp using primitive block. However I am not familiar with torch.. Is the inverse bilateral part of dpp implemented by:

1. unfolding every kernel to a vector like im2col
2. and then do the calculation
3. and finally reshape back ? Are there any plan about a pytorch implementation ? Thanks!