SimpleMind weights
Model weights used by the public SimpleMind knowledge-graph workflows. Each file is fetched, checked against its recorded sha256 and linked into a checkout by
python -m simplemind.kg assets fetch <workflow>
Files keep the layout of the checkout's weights/ folder. Downloading needs no account.
Licences
The licence differs by file; the repository setting above is "other" for that reason.
| File | Used by | Licence |
|---|---|---|
cxr/lungs.pth |
xray_chest_binary_segmentation |
BSD-3-Clause |
cxr/trachea.pth |
xray_chest_binary_segmentation |
BSD-3-Clause |
nnunet/kidney_lesion_ct.zip (one zip of the nnU-Net workspace folder) |
ct_abdomen_left_right_kidney_masks, ..._functional_kidney_masks, ..._functional_kidney_volume_measurements |
CC BY-NC-SA 4.0 |
Chest X-ray lung and trachea models (BSD-3-Clause)
Trained by the SimpleMind team. The maintainers have cleared them for sharing. The card does not list the training data. No training data is distributed here.
Kidney nnU-Net (CC BY-NC-SA 4.0)
Trained on the KiTS23 challenge data, whose images and segmentations are released under CC BY-NC-SA 4.0. Whether that licence reaches a trained model is unsettled, so the weights are released on the same terms to be conservative: non-commercial use, with attribution, share-alike. For commercial use, ask the KiTS organisers.
Attribution: Heller N, Isensee F, Tejpaul R, Wood A, Papanikolopoulos N, Weight C. The 2023 Kidney and Kidney Tumor Segmentation Challenge, MICCAI 2023. doi:10.5281/zenodo.7840134 (kits-challenge.org; github.com/neheller/kits23).
This is not legal advice.
Integrity
The sha256 of each file (the directory hash for the nnU-Net workspace) is recorded in
SimpleMind's plans_dir/kg/WEIGHTS.yaml, with the commit of this repository it was
uploaded at. kg assets fetch refuses a file whose hash does not match.
Not medical advice
Research use only. These models are not cleared for clinical use.