train.py can process successfully when i use horse2zebra data,but not using my own data

[vanhuyz]

train.py can process successfully when i use horse2zebra data,but not using my own data

[vanhuyz]

I'm not sure about 2-channel pictures but you can try to convert them to 3-channels using PIL, OpenCV or ImageMagick then train CycleGAN as usual 🤔

[IsDling]

@vanhuyz It‘s had processed successfully yesterday evening. Thank you for replying. It's very kind of you!
It had processed successfully yesterday evening. Thank you for replying. It's very kind of you!
I think it's better that it can do with a batch of photos in the process named "inference.py".

[gexiner]

@vanhuyz when I train my own data, it comes "Invalid JPEG data, size 333991", so what should I do to solve this problem? is that the size's problem?or....?

[XuJ1E]

@vanhuyz when I train my own data, it comes "Invalid JPEG data, size 333991", so what should I do to solve this problem? is that the size's problem?or....?

You can resize the weight and the height of image ，then try again！