---
tags:
  - medical-imaging
  - segmentation
  - nnu-net
  - airway
  - ct
---

# CaliTree weights

These are the trained weights for CaliTree, a solution to the ATM'26 (MICCAI 2026) airway tree modelling challenge.

Code: **[adinathdukre/CaliTree](https://github.com/adinathdukre/CaliTree)**

| folder | model | use |
|---|---|---|
| `track1/nnUNet_ckpts` | nnU-Net ResEnc-M, `nnUNetTrainerAirwayCB_clprec025`, 5 folds | Track 1 binary airway segmentation |
| `track2/nnUNet_ckpts` | nnU-Net ResEnc-M, `nnUNetTrainerAirwayCB`, 5 folds | Track 2 stage 1 airway mask |
| `track2/nnUNet_t2warm` | nnU-Net `3d_fullres_SAS`, `nnUNetTrainerT2Warm`, 5 folds | Track 2 stage 2 21-class voxel labelling |

Each track folder is used directly as `ATM_RESOURCES`:

```bash
huggingface-cli download adidukre/CaliTree --local-dir /path/to/weights
ATM_RESOURCES=/path/to/weights/track1 ATM_INPUT=/path/to/ct_folder ATM_OUTPUT=/path/to/output \
    python docker/T1/inference.py
```