Skip to content

Latest commit

 

History

History
96 lines (80 loc) · 3.15 KB

README.md

File metadata and controls

96 lines (80 loc) · 3.15 KB
Raw

Advance tracking with DeepSORT and YOLOv4

Key features

  • Select single object to track with Region of Interest (ROI)
  • Isolate tracking object as cropped frame
  • Plotting travelled distance and velocity (Applicable only if camera is static)
  • Tracjectory representation

Set up environment

# setup conda environment
$ conda env create -f environment.yml
$ conda activate deepsort_track 

# download weights
$ chmod +x download_weights.sh && ./download_weights.sh

# run unit tests to ensure all weights are in correct directories
# if all tests are passed, proceed to next step
$ pytest unit_tests.py

Run from terminal

usage: use "python tracking.py --help" for more information

Tracking with DeepSort

optional arguments:
  -h, --help            show this help message and exit
  -v VIDEO, --video VIDEO
                        Path to video file
  -t TARGET [TARGET ...], --target TARGET [TARGET ...]
                        Type of tracking target (person, car, etc)
  -c CONF, --conf CONF  Confident threshold (deafult=0.3)
  -n NMS, --nms NMS     NMS threshold (default=0.4)
  -d MCD, --mcd MCD     Max cosin distance (default=0.5)
  -f FREQ, --freq FREQ  Detection update frequency in second (default=2.0)
  -cl COLORS [COLORS ...], --colors COLORS [COLORS ...]
                        List of colors
  -s SAVE, --save SAVE  Save output video (True/False)


# run tracking with live representations
python tracking.py -v test_video/football.mp4 -t person

# run tracking and save result as videos
python tracking.py -v test_video/football.mp4 -t person -s True

Select tracking target

  • When the target which we desire to track is detected (bounding box around target), press "c" and click + drag left mouse to select the target
  • Press Enter to execute tracking or click left mouse to reselect the target

Citation

@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}
}

@misc{bochkovskiy2020yolov4,
     title={YOLOv4: Optimal Speed and Accuracy of Object Detection}, 
     author={Alexey Bochkovskiy and Chien-Yao Wang and Hong-Yuan Mark Liao},
     year={2020},
     eprint={2004.10934},
     archivePrefix={arXiv},
     primaryClass={cs.CV}
}

Credit