search

cscss / DBVSR

This repository is an official PyTorch implementation of the paper "Deep Blind Video Super-resolution"
37 stars 4 forks source link

question about degradation for test #2

Open duck9144 opened 3 years ago

duck9144 commented 3 years ago

Thanks for your excellent work! I have a question. Is the degradation for testing dataset Vid4 and SPMC bicubic downsampling? If it is, why the quantitative and visual results of other compared methods are not same as the original paper? Most methods are using bicubic downsampled trainning dataset.

cscss commented 2 years ago

The degradation for these datasets is bicubic downsampling. However, different calculation approaches will lead to different results. We re-generate these results and utilize the same calculation method for fair comparisons. So the quantitative and visual results of other compared methods are not the same as the original paper.