search

hhqweasd / G2R-ShadowNet

CVPR2021 From Shadow Generation to Shadow Removal
52 stars 12 forks source link

关于AISTD数据集：大家可以搜索关于“Shadow Removal via Shadow Image Decomposition”这篇文章的github，里面有AISTD数据集的下载链接 #13

Closed DoraemonPockets closed 1 year ago

DoraemonPockets commented 1 year ago 
          > Yes, Please use AISTD dataset

you mean evaluation use the AISTD dataset? where I can got the AISTD dataset? I can not find it, thanks.

Originally posted by @neverstoplearn in https://github.com/hhqweasd/G2R-ShadowNet/issues/6#issuecomment-1590517866

neverstoplearn commented 1 year ago 
          > Yes, Please use AISTD dataset

you mean evaluation use the AISTD dataset? where I can got the AISTD dataset? I can not find it, thanks.

Originally posted by @neverstoplearn in #6 (comment) 感谢，我试一下。

neverstoplearn commented 1 year ago 
          > Yes, Please use AISTD dataset

you mean evaluation use the AISTD dataset? where I can got the AISTD dataset? I can not find it, thanks.

Originally posted by @neverstoplearn in #6 (comment)

我下载了AISTD数据集， maskdir = 'test_mask'（作者提供的test_mask） shadowdir = 'AISTD' (作者提供的模型结果)
freedir = 'test_C_fixed_official'(AISTD数据集) 最终得到结果： PSNR(all,non-shadow,shadow): 30.522497 37.589435 32.146250 SSIM(all,non-shadow,shadow): 0.000000 0.000000 0.000000 PI-Lab(all,non-shadow,shadow): 3.862418 2.369646 10.006975 PP-Lab(all,non-shadow,shadow): 3.862418 2.325859 10.681674 请问我数据选择得有问题吗？谢谢。

DoraemonPockets commented 1 year ago

我在跑结果的时候shadowdir就是模型最后生成的去除阴影的结果图片，输出SSIM的那一段代码被注释掉了，你把前面的%去掉就可以得到SSIM了

neverstoplearn commented 1 year ago

我在跑结果的时候shadowdir就是模型最后生成的去除阴影的结果图片，输出SSIM的那一段代码被注释掉了，你把前面的%去掉就可以得到SSIM了 已经解决了