Hi I am trying to do multi-class segmentation (or binary class segmentation?) of brain MRI image. There are two ROI I have labeled and I only want to get two class results, But there also should be 'unknown' region in the image. Is it possible to treat this segmentation as binary segmentation while having this 'unknown' or 'unlabeled' region with semi-supervised training? How can I do semi-supervised UNET training to get segmentation of two class.
