YOLO_DeepSORT

Detect using YOLOv2 and track using DeepSORT

Reuqirement

tensorflow >= 1.0
cv2, numpy

Arch

YOLO_DeepSORT/
├── backup
	├── yolo_model_weights_file.weights
	...

├──cfg
	├──yolo_model_config_file.cfg
	├──yolo_model_class_type_file.data

├──resources
	├──networks
		├──tensorflow_model_weights_file.ckpt

├──0130 (it is a directory for frames saving)
  ├──01 (the serial id)
		├──det (for saving det.txt which contains all the detection informations)
		├──img1
			├──your_video_frames.jpg
			...

├──temp

├──your_yolo_lib_file.so

remember to modify *_path in deep_sort_app.py

How to run

python deep_sort_app.py  --display true --run_type instant
python deep_sort_app.py  --display true --run_type pre_computed

or

python3 deep_sort_app.py --display true --run_type instant
python3 deep_sort_app.py --display true --run_type pre_computed

see argparser() for more infomations about other arguments.

Have been tested under python 2.7/3.5/3.6 with tensorflow 1.2.0 (because I work with CUDA 8)

NOTICE

The tensorflow ckpt file can be download from here. See DeepSORT.

You can compile your own yolo lib file, but be sure that all the functions have the same usage as those in yolov2.py.

Citation

DeepSORT

If you find DeepSORT 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}
}

@inproceedings{Bewley2016_sort,
  author={Bewley, Alex and Ge, Zongyuan and Ott, Lionel and Ramos, Fabio and Upcroft, Ben},
  booktitle={2016 IEEE International Conference on Image Processing (ICIP)},
  title={Simple online and realtime tracking},
  year={2016},
  pages={3464-3468},
  doi={10.1109/ICIP.2016.7533003}
}

YOLOv2

Umm, he does not like this. YOLOv2