Some questions about the dataset for img to img translation task

[tianjilong123]

When I want to achieve the image-to-image task in my own datasets, there are some problems.

I noticed that the datasets for the img2img task in your paper is BraTS Brain MRI Dataset, which seems to be the .npy format.

If I want to use my own datasets, what's the structure of the data folder, and what format of the data(like .png or .npy).

Now, my datasets consists of the PNG image. What should I do to train the model successfully.

Hope for your reply.

[tianjilong123]

@mirthAI

[Sebastianjhx]

Apologies for the late reply and any confusion. I'll respond more promptly in the future.

To use a custom dataset:

1. Data folder structure: please refer to the README for the structure of data folders.
2. Custom dataset class: You'll need to build your own dataset class. Create your own dataset class by following the examples in the datasets/ folder. For instance, the LDFDCT and PMUB loaders demonstrate how to load PNG images.

Thanks