search

IDKiro / DehazeFormer

[IEEE TIP] Vision Transformers for Single Image Dehazing
MIT License
357 stars 34 forks source link

about the uploaded SOTS dataset #26

Closed diadestiny closed 1 year ago

diadestiny commented 1 year ago

Hi, thanks for your work! I have a simple but confusing question: the size of SOTS dataset you uploaded in Baiduyunpan link(data/RESIDE-IN/test.zip) is different from the raw official website version, I hope to know what the differences are. download from your upload image download from the raw official website image

the same problem with the other dataset(Reside-ITSv2) about your uploaded and raw versions.

diadestiny commented 1 year ago

the shape of the haze images is same, but the dataset size is different.

IDKiro commented 1 year ago

We excluded some small images from the training set of RESIDE OUT (because the resolution of these small images is smaller than the set patch size), and the names of the excluded files are recorded in the txt file. In addition, we use the ITS in RESIDE v0, which is consistent with most of the previous methods, as described in the paper.

IDKiro commented 1 year ago

I am very sorry, I found that you are asking about SOTS, I don't remember this part very well. I should have read all the images and saved them again, in order to avoid the difference of different image codecs reading JPEG format images. You can test it with the official test set and the results should be the same.

diadestiny commented 1 year ago

thanks, I will try it.