Xin Tao

I think your image has only one channel ? Then you can use `--model=gray` and modify a little the code to process only one channel.

Actually, you can check `blur.shape` to see what exactly it is.

I think maybe you need train longer time and pick a best model from last few checkpoints.

Do you mean the saved model from TensorFlow ? I think you saved intermediate values, such as gradients for each variable. In our released model, only parameters used in inference...

Sorry but this code was written using a very old TF version. You can try old version, maybe TF 1.4? It should work on newer version, but may need some...

You can test to see if the provided checkpoints work well on your data. In order to train on unpaired images, you need to search for other methods. I think...

There is a difference between released model and this table. We mentioned in the paper that in this table conv kernels are 3x3, in order for fast experiments. And in...

Yes. You can re-train use the checkpoints. You need to follow TensorFlow workflow, define new graph and initialize all variables. Then you can load our checkpoints. The variables in the...

That error means you new graph does not contain some operations which are in the checkpoint file. Maybe you can only load those required weights. I think there may be...

We set a learning strategy according to experience and did not try much. You may refer to other papers specifically for learning rates. They may help increase training speed.