Why my training stuck at the beginning

[Chen65010445]

The model was very fluid when I used the test images you provided. But when I used my own images for training, after typing 'python train.py' the model give me no response, nothing but flashing cursor. How did that happen? And how can I fix that?

[vanhuyz]

I haven't seen this issue before. Could you please give me more information about how to reproduce this issue? For example, your environment, what steps you did, your data information, which line in train.py where the program is stuck, etc.

[Chen65010445]

I replaced the default pictures you provide for test into my own pictures. I didn't change any code. I notice the default image size was 256256 in the code so I transform my own images into 256256. Then I followed the commands in 'readme'. Then when I use train.py there was no response as usual, even I waited a long time.

[Chen65010445]

Is it possible that I want to train with CT pictures instead of natural pictures like your default pictures?

[ghost]

I had the same problem using my own 512x512 dataset. The train start but stuck at the beginning without errors.

[Chen65010445]

Maybe you should read the code again, I found that the maximum pixel of picture in it is 256256. So maybe when we use a 512512 dataset for training it would not know what to do.......

[ghost]

Thank you. Do you know an CycleGan implementation working for bigger images, like 512x512 ?

[Chen65010445]

I still didn't find such implementation, so I choose to convert the picture into 256*256. You may find not too much different if you don't zoom it up. Sorry for can't help you....