About dataset

[dongzhue]

Hi Afifi,
Thanks for your awesome work!
I am curious about the generation of the dataset. In set 2, how did you generate the target white balance image from the image taken by mobile phones?
Is it also the same as the camera, using the raw format image to manually select the "ground-truth" white from the middle gray patches in the color rendition chart?

[mahmoudnafifi]

Hi, Thanks for your interest in the work. Set 2 includes several mobile phone cameras and we followed a similar process we did for set 1: raw images were rendered through the PS camera pipeline with different white-balance settings. The ground-truth images were generated by including a color chart and manually selecting a gray patch to correct the image in raw before rendering with accurate global white balance correction. The images were then cropped to remove the color chart from each scene.

[dongzhue]

Hi Afifi,
I have another question. For set2, do I have to use raw images with a color chart to make ground truth images? Can I use jpg images instead?
Is the color of the result obtained from the jpg image not so accurate? But I feel that the effect of the raw image on the mobile phone is not so good either. Maybe the difference between the raw and jpg on the mobile phone is not particularly large.
I would like to get your advice, thank you!