Skip to content

🐳 Docker Support

Aegear provides a ready-to-use Docker image to simplify deployment and training in cloud environments or on systems without a configured Python environment.

📦 Public Docker Image

The official Aegear Docker image is hosted on Docker Hub: 

ljubobratovicrelja/aegear:latest

Pull the image with: 

docker pull ljubobratovicrelja/aegear:latest

🚀 Running the Image

Launch a container with GPU support (if available): 

docker run --gpus all -it ljubobratovicrelja/aegear:latest

To mount a local directory for accessing datasets or saving models: 

docker run --gpus all -it -v /path/to/data:/workspace/data ljubobratovicrelja/aegear:latest

This will mount /path/to/data on your host to /workspace/data inside the container.

🛠 Custom Training with Docker

The public image includes all dependencies for running Aegear and training on the provided dataset. For custom datasets or workflows:

  1. Clone and Extend
    Clone the Aegear repository and modify the Dockerfile as needed:

    git clone https://github.com/ljubobratovicrelja/aegear.git
cd aegear
docker build -t aegear-custom .

  2. Custom Parameters
    When launching the container, you can specify:

  3. Which notebook to use (training_unet.ipynb or training_siamese.ipynb)
  4. Which branch of the repository (defaults to main)

Example:

docker run --gpus all -e NOTEBOOK=training_siamese.ipynb -e BRANCH=dev -it ljubobratovicrelja/aegear:latest

📌 Currently, this works for the main repository. Support for forks can be added easily by adjusting the Dockerfile’s repository source.

  1. Mount Your Data
    Mount your dataset and configuration files into the container using -v.

  2. Run Training Scripts
    Inside the container, launch Jupyter notebooks or Python scripts from the notebooks/ directory:

    jupyter notebook notebooks/training_unet.ipynb
jupyter notebook notebooks/training_siamese.ipynb

📖 Notes

  • GPU Support: Requires NVIDIA drivers and Docker with NVIDIA runtime (nvidia-docker2).
  • Cloud Deployment: Compatible with Google Cloud, AWS, and Azure GPU instances.
  • Extending: For additional libraries or dependencies, create a new Dockerfile based on ljubobratovicrelja/aegear:latest.

📦 Source Dockerfile

The Dockerfile used to build the public image is included in the repository at:

aegear/Dockerfile

This can serve as a starting point for creating your own custom images.