Thank you very much for releasing this code and the model weights. It is a very interesting and valuable project.
I have a question about the dataset structure that is required for validation and inference. Is it the case that for every label, a copy of the corresponding image must be saved? This seems very inefficient for datasets with multiple labels like Totalsegmentator as you would have many copies of a single CT in the different folders for heart, lung, etc.
Hi,
Thank you very much for releasing this code and the model weights. It is a very interesting and valuable project.
I have a question about the dataset structure that is required for validation and inference. Is it the case that for every label, a copy of the corresponding image must be saved? This seems very inefficient for datasets with multiple labels like Totalsegmentator as you would have many copies of a single CT in the different folders for heart, lung, etc.