3D Brain Tumor Segmentation: training files

Everything behind the trained models of Brain-Tumor-Segmentation-Mlops, so the results can be reproduced, inspected or built on. The ready-to-run ONNX models and demo data are in MohammadSakibHossain/brain-tumor-segmentation-mlops.

Path Contents
checkpoints/<model>/best.pt PyTorch weights of the best checkpoint (by validation Dice) of SegResNet, Swin UNETR and SegFormer3D
mlflow/mlflow.db MLflow tracking database: every training and test run with parameters, metrics and the model registry (brain-tumour-seg, alias @champion = SegResNet)
mlflow/mlruns.tar MLflow run files: training curves, example validation images, logged and registered models
preds/<model>/*.nii.gz test-set predictions (label maps) of each model
logs/ training, evaluation and export logs
backup.json source, date and SHA-256 checksum of every file

Use

In a clone of the GitHub repository:

python scripts/backup.py restore --repo MohammadSakibHossain/brain-tumor-segmentation-training --to .
mlflow ui --backend-store-uri sqlite:///mlflow.db        # browse all runs at http://localhost:5000

The restore checks every checksum and rewrites MLflow's stored folder paths for the new location. To load a checkpoint: ml.inference.load_model("output/runs/segresnet/best.pt", device).

Licence

Trained on the Medical Segmentation Decathlon Task01 BrainTumour (CC-BY-SA 4.0); these files are shared under the same licence. Research and education only: not a medical device.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support