inference on noisy images without masks

[tardigrade-10]

Hi, I am trying to predict clean images from noisy images directly without using masks, ultimately a predict.py file that take the input images only. how can this be done? and are you planning to work on it?

[GuoLanqing]

Thanks for your interest. You can use an external pre-trained shadow detector to generate the shadow mask. Or you can also re-train the shadow removal model and fix the shadow mask as the all-one matrix. We will try to include the mask-free version in our repo in the future.