Run the container

1. Create a directory for model caching (only necessary once):
   mkdir -p ~/.torch/fvcore_cache
2. Change to the docker directory of this repository:
   cd docker
3. Create environment variable file to store the user and group ids (only necessary once):
   echo UID=$(id -u)>.env && echo GID=$(id -g)>>.env
4. Adjust parameters (e.g. volumes) in docker-compose.yml.
5. Run the docker container:
   docker-compose run detectron2

Rebuild the container

Rebuilding the container is only necessary when Dockerfile has been changed. The initial build is done automatically.

1. Change to the docker directory of this repository:
   cd docker
2. Create environment variable file to store the user and group ids (only necessary once):
   echo UID=$(id -u)>.env && echo GID=$(id -g)>>.env
3. Trigger the build:
   docker-compose build detectron2

Install new dependencies

Persistent

Add the dependencies at the end of Dockerfile.

Example:

...
# Customization
USER root
RUN apt-get update && apt-get install -y \
  nano vim emacs
RUN pip install pandas
USER appuser

Temporary

Use sudo (with your usual credentials) inside of the container. Changes will be lost, when the container is restarted.

Example:
sudo apt-get update
sudo apt-get install nano vim emacs sudo pip install pandas