Some question about your pretrain model

[Jingwei-Liao]

Could you please share your pre-train model ? We want to compare our method with yours

[xw-hu]

The original pre-trained model was lost. You can train the model by yourself, and it is very fast.

[Jingwei-Liao]

hi bro, we retrain some shadow detection model on your dataset, but we find we cann't get the performance you proposed on your paper , we find the BER value of DSD and BDRAR is 1-2 higher than the data you proposed in your paper. We want to ask Have you modified other parameters besides the input image size, such as batchsize,training iter numbers, etc in DSDNet and BDRAR

[xw-hu]

If you follow the Readme to train the model and have the same versions of PyTorch, and others, it is easy to obtain similar results.

[xw-hu]

For DSD and BDRAR, please check the size of the input images. The original sizes used in their papers are smaller.

[Jingwei-Liao]

thank you very much