Data set processing

[1shifeng]

Hello, take HS-ERGB as an example, the grayscale in the article converts the data set into gray and then trains and tests it, and where is the big fuzzy data set?

[XiangZ-0]

Hi, we also train GEM models on the RGB color space on the HS-ERGB dataset. For the gray metrics in Tab. 1 and 2, we simply convert the RGB outputs to grayscale to compute PSNR and SSIM, so you can see both color and gray metrics in Tab. 1 and 2.

As for the dataset, this link contains the pre-processed HS-ERGB dataset with normal blur. We will synthesize large blur by averaging consecutive blurry images during training (please refer to the code here). Hope this helps.

[1shifeng]

OK,thank you for your reply.