This repository contains a moded version of PyTorch YOLOv3 (https://github.com/ultralytics/yolov3). It filters out every detection that is not a person. The detections of persons are then passed to a Deep Sort algorithm (https://github.com/ZQPei/deep_sort_pytorch) which tracks the persons. The reason behind the fact that it just tracks persons is that the deep association metric is trained on a person ONLY datatset.

The implementation is based on two papers:

• Simple Online and Realtime Tracking with a Deep Association Metric https://arxiv.org/abs/1703.07402
• YOLOv3: An Incremental Improvement https://arxiv.org/abs/1804.02767

Python 3.7 or later with all of the pip install -U -r requirements.txt packages including:

• torch >= 1.3
• opencv-python
• Pillow

All dependencies are included in the associated docker images. Docker requirements are:

• nvidia-docker
• Nvidia Driver Version >= 440.44

Github block pushes of files larger than 100 MB (https://help.github.com/en/github/managing-large-files/conditions-for-large-files). Hence the yolo weights needs to be stored somewhere else. When you run tracker.py you will get an exceptions telling you that the yolov3 weight are missing and a link to download them from. Place the downlaoded .pt file under yolov3/weights/. The weights for deep sort are already in this repo. They can be found under deep_sort/deep/checkpoint/.

track.py runs tracking on any video source:

python3 track.py --source ...

• Video: --source file.mp4
• Webcam: --source 0
• RTSP stream: --source rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa
• HTTP stream: --source http://wmccpinetop.axiscam.net/mjpg/video.mjpg

If you find this project useful in your research, please consider cite:

@misc{yolov3-deepsort,
    title={Real-time multi-camera multi-object tracker using YOLOv3 and DeepSORT},
    author={Mikel Broström},
    howpublished = {\url{https://github.com/mikel-brostrom/Yolov3_DeepSort_Pytorch}},
    year={2019}
}

For more detailed information about the algorithms and their corresponding lisences used in this project access their official github implementations.