LitDet

LitDet is a domain-agnostic AutoML framework to accelerate the development of 2D object detection models 🚀⚡🔥.

• ML Workflows: Handles model pre-training, fine-tuning or re-training at scale.
• Key Features: Provides a high-level CLI and python API for seamless hardware support (CPU/GPU), COCO-formatted dataset integration, and built-in experiment tracking.
• Tech Stack: Built on the Lightning framework, it leverages PyTorch and Hydra for flexible, configuration-driven deep learning workflows.

Quickstart

You will need at least python 3.10 installed on your machine. For efficient training, it is recommended to have a recent GPU driver (e.g. NVIDIA 525.147.05). It also works on Mac with MPS support.

1. First install the package (best in a python venv).

   python -m pip install "litdet[extras]"
   

2. Train a Faster-RCNN on your dataset annotated in COCO format at /path/to/your/dir/coco_dataset_name:

   light-train task/model=faster_rcnn paths.data_dir=/path/to/your/dir data.data_name=coco_dataset_name task.model.num_classes=10 trainer.max_epochs=100
   

Replace the 10 by the number of classes your model should predict.

You can find configuration examples in the /examples directory.

To build your own configuration file instead, use our cookiecutter:

cookiecutter https://gitlab.kitware.com/litdet/litdet.git --directory "cookiecutter-litdet"

For more information on the usage, configuration, API documentation, and getting started guides, refer to our online documentation.

Developers

See CONTRIBUTING for developers instructions such as code practices, or the review process. Check also our advanced usage guide, and debugging process.

Citation

You can cite LitDet with the following:

@article{litdet2026tetrel,
  title={LitDet: An AutoML Tool for Accelerating 2D Object Detection Experiments},
  author={Tetrel, Loic and Checchin, Thomas and Pagnier, Louis},
  journal={Journal of Open Source Software},
  year={2026}
}

Model Zoo

The LitDet model zoo serves as a centralized hub for users and developers to discover models.

You can explore the complete list of available pre-trained models directly in our GitLab repository.

Acknowledgements