Training instability
Opened this issue · 0 comments
tawsinDOTuddin commented
I am working on old video restoration as my masters thesis. And so far I have been following deepremaster which is the most relevant one for old video restoration. However, due to lackings of guided structural artifacts learning (using only L1 loss), it was giving me unstable learning like it was not paying attention to the artifacts. Do you have any idea of this?