SAMannot

A Segment Anything based video annotation tool.

For efficient keyframe annotation and data propagation.

GitHub Paper

Gergely Dinya1, András Gelencsér1, Krisztina Kupán2, Clemens Küpper2, Kristóf Karacs3,
Anna Gelencsér-Horváth1,3*

1 Faculty of Informatics, Eötvös Loránd University, Budapest, Hungary
2 Max Planck Institute for Biological Intelligence, Seewiesen, Germany
3 Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest, Hungary

* Corresponding author: gha@itk.ppke.hu

Citation

@misc{samannot,
              title={SAMannot: A Memory-Efficient, Local, Open-source Framework for Interactive Video Instance Segmentation based on SAM2},
              author={Gergely Dinya and Andr{\'a}s Gelencs{\'e}r and Krisztina Kup{\'a}n and Clemens K{\"u}pper and Krist{\'o}f Karacs and Anna Gelencs{\'e}r-Horv{\'a}th},
              year={2026},
              eprint={2601.11301},
              archivePrefix={arXiv},
              primaryClass={cs.CV},
              url={https://arxiv.org/abs/2601.11301},
            }

About

SAMannot is a versitile video annotation tool built on top of Meta's Segment Anything Model (SAM2). It helps you create high-quality segmentation masks across video frames with minimal user interaction.

Features

Getting Started

For full details, see the SAMAnnot repository. A minimal example workflow is shown below.

Clone the repository

git clone https://github.com/gergelydinya/SAMAnnot.git
cd SAMAnnot
conda create -n samannot python=3.10 -y
conda activate samannot

Installation (Linux)

pip install -r requirements.txt
cd sam2
pip install -e .
cd..
pip install --index-url https://download.pytorch.org/whl/cu121 torch torchvision torchaudio
cd ../checkpoints
./download_chckpts.sh

Run the tool

 conda activate 
python main.py