What is the exact input shape of model?

[NicolasLaube]

I integrated/adapted your code to my pipeline. However, when I train the model, I get some strange results. I think that it comes from the input matrices of cartoons (edged/gray/normal)...

What are the expected cartoon and picture input shapes/order of channels ?

[ptran1203]

• Training images are read by cv2, default color space is RGB, but I reversed it to BGR.
• There is no particular input shape, it can be any, my dataset use 256x256 and you can inference on any shape
• What dataset do you use in training?

[NicolasLaube]

Thank you for your quick response ! We are training the model on a custom cartoons dataset (more Pixar like images).

[NicolasLaube]

I still have a last question, is the shape (3, im_size_height, im_size_width) ?

[ptran1203]

I still have a last question, is the shape (3, im_size_height, im_size_width) ?

Yes, in PyTorch channel first order is default

If you train on custom data, it's hard to say but you may need to tune some params (e.g loss weight, ...)

[NicolasLaube]

Ok, thank you !