Giters

google-research / maxim

[CVPR 2022 Oral] Official repository for "MAXIM: Multi-Axis MLP for Image Processing". SOTA for denoising, deblurring, deraining, dehazing, and enhancement.

Home Page:https://arxiv.org/abs/2201.02973

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

MIT 5K Data pre-processing

mv-lab opened this issue · comments

commented

Dear authors,

you mention a random partition for train/test of the MIT 5K dataset, instead of the "popular" adopted test set splited by DeepUPE. How do you generate the input RGBs from the provided DNGs? (i.e. using lightroom software or other's method preprocessing?)

can you provide the ids of the validation set, so we can compare fairly?

Thank you.

commented

Hi @mv-lab, thanks for bringing up this issue. Yes, there exist different versions of MIT 5K. I'm not following this field, so we just downloaded the preprocessed dataset provided in UEGAN for convenience. We use the first 4500 and last 500 images for training and testing, following MIRNet.