fidoptMedical SAM3
SAM 3 fine-tuned on 33 medical datasets, its authors' 2D model prompted by text alone, no box or point: the structure the authors' prompt names, the mask of highest score above 0.1, as their evaluation script runs it.
SegmentationGPUversion Medical-SAM3 b610242, checkpoint_2D.pt of huggingface.co/ChongCong/Medical-SAM3 at 6ea30a2 (sha256 f7fb9f20; its model's storages 261296b5)fetus, head and neck · USSAM License (code); weights: MIT on the model card, a fine-tune of SAM 3's weights (SAM License)
Run it
With its KonfAI app, at the benchmark's settings (the fidopt app repositories are private for now: a Hugging Face token with access opens them):
pip install konfai-apps
| prompt | Command | |
|---|---|---|
| fetal head | Get the app | konfai-apps infer fideus-labs/fidopt-segmentation:MedicalSAM3 -i image.nii.gz -o out --gpu 0 --set=structure=fetal_head |
| thyroid nodule | Get the app | konfai-apps infer fideus-labs/fidopt-segmentation:MedicalSAM3 -i image.nii.gz -o out --gpu 0 --set=structure=thyroid_nodule |
Results
| Dataset | prompt | # | Read by | Score | Failed | Time s |
|---|---|---|---|---|---|---|
| DDTI, thyroid nodules outlined on ultrasound | thyroid nodule | Dice | 0.393 | 0 | ||
| HC18, the fetal head as a sonographer measured it | fetal head | Dice | 0.896 | 1 |
- Reference
- Jiang et al., arXiv 2601.10880, 2026