The Model gives opposite results

[shubham12tomar]

The model is not detecting the actual shadow.
It only detect dark colors as shadow in the input images during Testing using test.py and ST-CGAN_G2_1500.pth weights.
I wondered, if this model actually work as shown in your GitHub Repository. But I really appreciate your work.

I was run your code on Google Colab.

[Param-Raval]

Hello!

Yes, the model does remove shadows to an extent on the images we tested. In fact, this particular model was originally proposed keeping in mind the issue of detecting dark colors that you mentioned. Yet, I would suggest that you look through the original paper and its code and implement from there since we too started from their implementation.

Thank you!