FVL2020 / ICCV-2023-MB-TaylorFormer

84 stars 9 forks source link

Evaluation Result on Dense-Haze #13

Open JJJJJoshua opened 5 months ago

JJJJJoshua commented 5 months ago

Hello, I try to evaluate the pretrained model (i.e., densehaze-MB-TaylorFormer-L.pth) on Dense-Haze dataset, and I end up getting images attached, which is not very good. I'm not sure if I did something wrong or the results are correct. Could you tell me if they are correct?

Anyways, the method proposed is very solid and the results on ITS&OTS are satisfying. If possible, could you share some ideas on the performance on Dense-Haze? I think it might be that transformers need large datasets to perform well. Thank you so much! 51_hazy 52_hazy

Auorui commented 4 months ago

I feel that the effect here is not as good as I expected, which may be related to the dataset itself 54_hazy 54_GT