Open-VOT is a lightweight library of visual object tracking for research purpose. It aims to provide a uniform interface for different datasets, a full set of models and evaluation metrics, as well as examples to reproduce (near) state-of-the-art results.
Currently two trackers are implemented: SiamFC and GOTURN. We aim to include more state-of-the-art trackers. Please comment if you have any recommendations.
Currently only Python3 is supported.
Install PyTorch (version >= 0.4.0) and other dependencies:
pip install h5py scipy matplotlib tensorboardXTracking with pretrained models
Download pretrained models from here, then execute the following command:
python examples/goturn --phase testTraining
python examples/siamfc --phase trainThis command trains SiamFC tracker on the VOT2017 dataset. Note VOT2017 might be too tiny to train deep networks.
Please check out more examples in the examples folder. Or, alternatively, you could find out more concised examples in the tests/trackers folder.