This repository contains code to run DeepSORT on iWilcam 2021 camera trap image sequences.
Please refer to the original DeepSORT repository for more information about Deep SORT or see the arXiv preprint.
To extract features using EfficientNet-B2, use the script mot/generate_features.py from our main iWildcam repository. We kept DeepSORT code on a separate repository to avoid GPLv3 licensing conflicts.
To track animals with DeepSORT use the script track_iwildcam.py:
python track_iwildcam.py --test_info_json=PATH_TO_BE_CONFIGURED/iwildcam2021_test_information.json
--features_json=PATH_TO_BE_CONFIGURED/efficientnet_b2_crop_25mai_features.json
--tracks_file=PATH_TO_BE_CONFIGURED/efficientnet_b2_crop_25mai_tracks.jsonFinally, to classify tracks and generate a submission, go back to from our main iWildcam repository and use the script classification/predict_track.py.
If you find this repo useful in your research, please consider citing the following papers:
@inproceedings{Wojke2017simple,
title={Simple Online and Realtime Tracking with a Deep Association Metric},
author={Wojke, Nicolai and Bewley, Alex and Paulus, Dietrich},
booktitle={2017 IEEE International Conference on Image Processing (ICIP)},
year={2017},
pages={3645--3649},
organization={IEEE},
doi={10.1109/ICIP.2017.8296962}
}
@inproceedings{Wojke2018deep,
title={Deep Cosine Metric Learning for Person Re-identification},
author={Wojke, Nicolai and Bewley, Alex},
booktitle={2018 IEEE Winter Conference on Applications of Computer Vision (WACV)},
year={2018},
pages={748--756},
organization={IEEE},
doi={10.1109/WACV.2018.00087}
}