Instructions to use VBoussot/ImpactReg with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- KonfAI
How to use VBoussot/ImpactReg with KonfAI:
pip install "konfai[imaging]" konfai-apps
# List the apps this repository holds konfai-apps list VBoussot/ImpactReg # Run one of them on an image konfai-apps infer VBoussot/ImpactReg:<app> -i image.nii.gz -o Output
- Notebooks
- Google Colab
- Kaggle
Download ConvexAdam_IMPACT_MRCT/app.json from VBoussot/ImpactReg: direct link, hf CLI and curl.
- Browser
- Download file 4.72 kB
-
https://huggingface.co/VBoussot/ImpactReg/resolve/main/ConvexAdam_IMPACT_MRCT/app.json
- Command line
-
hf download hf://VBoussot/ImpactReg/ConvexAdam_IMPACT_MRCT/app.json
-
curl -L -o app.json https://huggingface.co/VBoussot/ImpactReg/resolve/main/ConvexAdam_IMPACT_MRCT/app.json
4.72 kB
| { | |
| "display_name": "ConvexAdam, MIND + TotalSegmentator MR segmentation", | |
| "short_description": "ConvexAdam's coarse search on MIND and the segmentation head of TotalSegmentator MR (TS/M730), for MR/CT.", | |
| "description": "ConvexAdam's coupled-convex coarse search on MIND and the segmentation head of TotalSegmentator MR (TS/M730), for MR/CT: the one-hot map of its 30 structures, compared by squared differences like MIND, after the affine pre-alignment on mutual information. No Adam refinement: on MR/CT pairs the refinement folded 1 to 2 % of the body and cost the head-and-neck pair about 0.2 of Dice, where the coarse field alone reached a Dice of 0.665, 0.599 and 0.686 on the Learn2Reg abdomen, SynthRAD abdomen and SynthRAD head-and-neck MR/CT pairs, with no fold.", | |
| "task": "registration", | |
| "tta": 0, | |
| "mc_dropout": 0, | |
| "models": [], | |
| "inputs": { | |
| "Fixed": { | |
| "display_name": "Fixed image", | |
| "volume_type": "VOLUME", | |
| "required": true | |
| }, | |
| "Moving": { | |
| "display_name": "Moving image", | |
| "volume_type": "VOLUME", | |
| "required": true | |
| }, | |
| "FixedMask": { | |
| "display_name": "Fixed mask (optional)", | |
| "volume_type": "SEGMENTATION", | |
| "required": false, | |
| "default": "ones" | |
| }, | |
| "MovingMask": { | |
| "display_name": "Moving mask (optional)", | |
| "volume_type": "SEGMENTATION", | |
| "required": false, | |
| "default": "ones" | |
| } | |
| }, | |
| "outputs": { | |
| "DisplacementField": { | |
| "display_name": "Displacement field", | |
| "volume_type": "VOLUME", | |
| "required": true | |
| } | |
| }, | |
| "inputs_evaluations": { | |
| "Image": { | |
| "Evaluation_with_images.yml": { | |
| "FixedImage": { | |
| "display_name": "Fixed image", | |
| "volume_type": "VOLUME", | |
| "required": true | |
| }, | |
| "MovingImage": { | |
| "display_name": "Moving image", | |
| "volume_type": "VOLUME", | |
| "required": true | |
| }, | |
| "Mask": { | |
| "display_name": "Evaluation mask", | |
| "volume_type": "SEGMENTATION", | |
| "required": false | |
| } | |
| } | |
| }, | |
| "Segmentation": { | |
| "Evaluation_with_seg.yml": { | |
| "FixedSeg": { | |
| "display_name": "Fixed segmentation", | |
| "volume_type": "SEGMENTATION", | |
| "required": true | |
| }, | |
| "MovingSeg": { | |
| "display_name": "Moving segmentation", | |
| "volume_type": "SEGMENTATION", | |
| "required": true | |
| } | |
| } | |
| }, | |
| "Landmarks": { | |
| "Evaluation_with_fid.yml": { | |
| "FixedFid": { | |
| "display_name": "Fixed landmarks", | |
| "volume_type": "FIDUCIALS", | |
| "required": true | |
| }, | |
| "MovingFid": { | |
| "display_name": "Moving landmarks", | |
| "volume_type": "FIDUCIALS", | |
| "required": true | |
| } | |
| } | |
| } | |
| }, | |
| "vram_bytes_per_voxel": 1150, | |
| "tiling": { | |
| "global": "Prediction.yml" | |
| }, | |
| "ram_bytes_per_voxel": 2050, | |
| "model_files": [ | |
| { | |
| "repo_id": "VBoussot/impact-torchscript-models", | |
| "revision": "47ffad660e67cebbac8d74aecf0fa181c5dac042", | |
| "filename": "MIND/R1D2_3D.pt" | |
| }, | |
| { | |
| "repo_id": "VBoussot/impact-torchscript-models", | |
| "revision": "47ffad660e67cebbac8d74aecf0fa181c5dac042", | |
| "filename": "TS/M730.pt" | |
| } | |
| ], | |
| "sample": { | |
| "label": "An abdominal MR (fixed) and CT (moving) of one patient, with ROI masks and four organs labelled (fidopt L2R_AbdomenMRCT case_0001, Learn2Reg 2021; CT: TCIA TCGA, CC BY 3.0; MR: CHAOS, CC BY-NC-SA 4.0)", | |
| "inputs": { | |
| "Fixed": "L2R_AbdomenMRCT/development/case_0001/fixed.ome.zarr", | |
| "Moving": "L2R_AbdomenMRCT/development/case_0001/moving.ome.zarr", | |
| "FixedMask": "L2R_AbdomenMRCT/development/case_0001/fixed_mask.ome.zarr", | |
| "MovingMask": "L2R_AbdomenMRCT/development/case_0001/moving_mask.ome.zarr" | |
| }, | |
| "references": { | |
| "FixedSeg": "L2R_AbdomenMRCT/development/case_0001/truth/fixed_labels.ome.zarr", | |
| "MovingSeg": "L2R_AbdomenMRCT/development/case_0001/truth/moving_labels.ome.zarr" | |
| } | |
| } | |
| } | |