jhuguetn/shivai-runtime repository overview

⁠SHiVAi runtime Docker image

Portable CPU-only Docker runtime for reproducible SHiVAi execution.

• SHiVAi project: https://github.com/pboutinaud/SHiVAi⁠

Find the image in Docker Hub here⁠.

⁠Components

• Debian 11 Bullseye
• Python 3.11.13
• ANTs 2.4.3
• niimath
• dcm2niix
• TensorFlow 2.17.0
• Keras 3.6.0
• SHiVAi pinned commit/version: b57c49175861b9a434e6b8bc3b0491faf4a0aac9

⁠Usage

The container entrypoint is shiva.

Show help:

docker run --rm -it jhuguetn/shivai-runtime:latest --help

Example PVS run:

mkdir -p ./data/out

docker run --rm -it \
  -v $(pwd)/config.yml:/config.yml:ro \
  -v $(pwd)/models:/models:ro \
  -v $(pwd)/data:/data \
  jhuguetn/shivai-runtime:latest \
  --in /data/in \
  --out /data/out \
  --input_type standard \
  --prediction PVS \
  --brain_seg fs_precomp \
  --use_cpu \
  --config /config.yml

⁠License

This repository contains Docker build infrastructure for SHiVAi runtime execution.

The distributed image installs SHiVAi, which remains governed by its upstream license:

• SHiVAi pipeline and repository content: GNU Affero General Public License v3.0 or later (AGPLv3+)
• SHiVAi trained model weights: CC BY-NC-SA 4.0

This image does not include model weights. Models and configuration files must be mounted at runtime.

⁠Notes

• CPU-only image
• Generic SHiVAi runtime
• Models and configuration files should be mounted at runtime
• Intended for reproducible research workflows as a parent image for future workflow-specific images

⁠Credits

Jordi Huguet (BarcelonaBeta Brain Research Center⁠)