Installation

Python pip

Requirements

You will need at least python 3.10 and a recent GPU driver (e.g. NVIDIA).

If using NVIDIA, we require at least 525.147.05 for CUDA 12 compatibility.

Note

We recommend using a python venv for a pip installation.

python -m venv .litdet-venv
source .litdet-venv/bin/activate

From public wheels

python -m pip install --upgrade pip
pip install litdet[extras]

From source

You can also build from source if needed.

git clone https://gitlab.kitware.com/litdet/litdet
python -m pip install --upgrade pip
pip install .

Docker

We support Docker to have a fully reproducible software environment.

Requirements

If you have docker installed on your machine, you can run the project directly by pulling our image or building it. You will require:

Build the image

You can also build the image yourself:

sudo docker build -t litdet:latest .

Apptainer

Appptainer allows users to run high-performance computing applications without requiring root privileges or compromising host system security. For more information on the difference between apptainer` and singularity, check this thread.

Requirements

You will require:

Build the image

The container is available for runtime, you can simply build it on your machine from our docker registry. First, make sure you have an account on our gitlab (https://gitlab.kitware.com/) and log-in into our docker registry:

apptainer registry login -u <USER> -p <TOKEN> docker://gitlab.kitware.com:4567

The USERNAME variable is the one from your gitlab account. You also need to create a new personnal token access TOKEN` with registry read in settings/access_tokens and Project access tokens section.

Then simply pull/build the singularity image:

apptainer pull litdet-latest.sif docker://gitlab.kitware.com:4567/litdet/litdet:latest