search

Royalvice / DocDiff

ACM Multimedia 2023: DocDiff: Document Enhancement via Residual Diffusion Models. Also contains 1597 red seals in Chinese scenes, along with their corresponding binary masks.
https://www.aibupt.com/
MIT License
234 stars 24 forks source link

图片降噪问题 #22

Closed leung-yaya closed 12 months ago

leung-yaya commented 12 months ago

大佬好，我这边用了这张图片进行降噪，发现好像比之前还模糊的。 原图： 608_608 结果： ![Uploading finalImgs.png…]()

leung-yaya commented 12 months ago

finalImgs

Royalvice commented 12 months ago

预训练权重吗？

leung-yaya commented 12 months ago

是的，用的是预训练权重

Royalvice commented 12 months ago

预训练的那个数据集的样本多样性比较小，像这种多种噪声的图像效果不行得自己构建数据集，重新训练

leung-yaya commented 12 months ago

感谢大佬解答

Royalvice commented 12 months ago

感谢大佬解答

不用客气