zejinwang / Blind2Unblind

This is an official implementation of the CVPR2022 paper "Blind2Unblind: Self-Supervised Image Denoising with Visible Blind Spots".
https://arxiv.org/abs/2203.06967
122 stars 14 forks source link

代码使用请教 #9

Closed Mrhard1999 closed 1 year ago

Mrhard1999 commented 2 years ago

请问您这个代码有专用的数据集吗?如果有能否提供数据集链接?或者说用含有噪声的图片就可完成训练?

zejinwang commented 1 year ago

用噪声数据集即可完成训练,只需要保证数据集的噪声满足泊松高斯分布。如果要用公开数据集进行训练,可以使用SIDD数据集,链接为:https://www.eecs.yorku.ca/~kamel/sidd/benchmark.php