Open IPNUISTlegal opened 6 years ago
You should download NYU dataset.Then you can load the nyu_depth_data_labeled by MATLAB.After that,you can check out the variables about the .mat, you will find 2200+images. Next, you could read this images and depth.mat and save it by yourself. I have trained the model sucessfully. But as for the correction model, I have no idea.
https://github.com/kskin/WaterGAN/issues/3 The code seems to fail without depth data(*.mat).