Can I get balance sample in case level?

[Bigsealion]

Hi:
In my datasets, most of the parts have label 1, and small parts have labels 1 and 2, which creates a huge imbalance in training.
So, I want to get a balanced sample (subject with label 1 and subject with labels 1&2 have the same number) in each epoch. I have known Torchio can get a balanced sample in patch level by LabelSampler, but can Torchio create a balanced sample in case level easily?

Thanks!

[fepegar]

This is not implemented at the moment, but I think you can use a PyTorch Sampler for this.

[Bigsealion]

Thanks for your reply! It's really helpful!