splinter21
splinter21
> > I got it. > > If the goal is to upscale a natural video to larger resolution, average pooling is ok. 0.5x(avg)->1x(train) vs 1x->2x(no ground truth)(test). > >...
https://github.com/clavichord93/diagonalwise-refactorization-pytorch/issues/1#issuecomment-388640922 It's said that it's optimized in the latsest version of pytorch, maybe this work is helpful for old pytorch version users?
When there is a hole in the high frequency spectrum, I have experienced nan loss.
Up to now, no one has compared these two methods.