search

swz30 / Restormer

[CVPR 2022--Oral] Restormer: Efficient Transformer for High-Resolution Image Restoration. SOTA for motion deblurring, image deraining, denoising (Gaussian/real data), and defocus deblurring.
Other
1.74k stars 233 forks source link

Questions about the quantitative results of other methods? #76

Closed C-water closed 1 year ago

C-water commented 1 year ago

Hi, How are the quantitative results calculated for the other methods in Restormer Table 1? Are you quoting their results directly or are you retraining them?

Looking forward to your reply. Thank you!

adityac8 commented 1 year ago

For Image Deraining, we follow the same convention as MSPFN and utilize the Table provided in their paper.

C-water commented 1 year ago

Thank you, I have downloaded others quantitative results from MSPFN. Besides, the Restormer only uses Rain100L as validation in the training stage? Or uses (rain100L+rain100H+test100+test1200+test2800) datasets as validation in the training stage?

Thank you again, and looking forward to your reply.

adityac8 commented 1 year ago

We cross validate our trainind on Rain100L dataset. You can change it to use another dataset here https://github.com/swz30/Restormer/blob/733ceb2e5cdab8074d59a8d440c645a11b5dd334/Deraining/Options/Deraining_Restormer.yml#L46-L47