search

LoSealL / VideoSuperResolution

A collection of state-of-the-art video or single-image super-resolution architectures, reimplemented in tensorflow.
MIT License
1.61k stars 295 forks source link

FRVSR train problems #84

Closed lius-xjtu closed 4 years ago

lius-xjtu commented 5 years ago

Can you tell me which dataset your model is training on, and batchsize?

lius-xjtu commented 5 years ago

I trained with vimeo data and found the color is wrong on the test image.

LoSealL commented 5 years ago

I have trained with VIMEO too, is the wrong color shown on the validation board, or on inferred outputs?