Generated original images is black

[dhu5-utk]

Hello,

Thanks for your uploading. I follow the instruction and successfully run the code. However, There is a problem with the result. The original image is totally black and thus the predicted image cannot make sense. BTW, the true mask image is correct. I appreciate your help,

Best, Da Hu

[yiskw713]

Hi,

Thank you for visiting and looking at my repository.

There were some mistakes:

• the values of std and mean in dataset.py and predict.py were wrong. This is because I had used PIL images, where pixel values range from 0 to 255. Now I save images from Tensor directly, so pixel values need to be between 0 and 1
• In predict.py, save_image(x/255, args.result_path + '/' + 'original_images_with_' + args.model + '.jpg') => save_image(x, args.result_path + '/' + 'original_images_with_' + args.model + '.jpg') This caused from the same reason.

I fixed them, so please try to run predict.py . If you used the same values of std and mean as my code, you may have to run train.py once again.

I'm sorry for inconvenience. This is my first implementation of deep learning with pytorch. So I never expected that someone checked my repository. If you have further questions, please do not hesitate to ask. Thanks.

Yuchi