Segment Anything model(SAM) on the RADIOMICS dataset


We have been playing around with the Segment Anything model(SAM) released last week by Meta.

Our initial experiments show a decent performance on samples from the Radiomics dataset. We however see a strong need to validate and correct the labeled data with a human expert in the loop process. Do you have any recommendations or experiences to share around this?

There’s a discussion currently going on this topic here:


Thank you! Pretty interesting developments!