search

Ree1s / IDM

289 stars 22 forks source link

It seems that the part of SR on general images for DIV2K dataset has not been realized ?? #26

Open Sue-syx opened 11 months ago

Sue-syx commented 11 months ago

According to the current code for config file and dataset, 'lr_dataroot' and 'hr_dataroot' keys in config files are not used for creating dataset. And current dataset-related codes, such as 'init.py' and 'LRHR_dataset.py', are only fitted to FFHQ and CelebAHQ dataset. Besides, the cropped patches of lr_size 16 and hr_size 128 for DIV2K confuse me. In this way, I wonder whether the part of SR on general images for DIV2K dataset has been realized.

Ree1s commented 11 months ago

Thank you for your interest. We train on DIV2K with random 40X40 lr patch size and 160X160 hr patch size. However, please note that the method for DIV2K data processing is likely to be consistent with other research like LAR-SR. We have not made it publicly available at this time.

Lipurple commented 11 months ago

@Ree1s Thank you for your reply. But I also want to know why in the df2k_liifsr3_x4.json, the resolution of lr and hr is 16 and 128? If this isn't the true config, when the true config will be released? Further, I am confused about how to use the df2k_gen.pth, I try many resolutions but the quality of images seem poor.

prokaryote-hgy commented 2 months ago

Thank you for your interest. We train on DIV2K with random 40X40 lr patch size and 160X160 hr patch size. However, please note that the method for DIV2K data processing is likely to be consistent with other research like LAR-SR. We have not made it publicly available at this time.

Hi, do you crop 40X40 and 160x160 patches from the original image, or resize the whole image to 40x40 and 160x160?