otre99 / tracking_with_ocv

Object tracking with python

This program tracks multiple objects in a video

Dependences

It only depends on OpenCV 4.5. You can use pip3 to install the dependencies:

pip3 install -r ./requirements.txt

Run the program

Run this to see the input options:

python3 main.py --help

usage: This program tracks multiple objects in a video [-h] [--video VIDEO] [--json JSON] [--method {CSRT,KCF,DaSiamRPN}] [--output OUTPUT] [--display]

options:
  -h, --help            show this help message and exit
  --video VIDEO         Input video
  --json JSON           JSON file containing initial bboxes
  --method {CSRT,KCF,DaSiamRPN}
                        Tracking method
  --output OUTPUT       Output video
  --display             Show results in the screen

The program expects a video (--video ) and a JSON file (--json) containing the starting position coordinates ( [x,y,w,h] rectangle) of each object you want to track (see this example).

Examples:

• Run with default parameters. It will use the video and JSON file inside folder development_assets:

  python3 main.py
  

• Show results on the screen:

  python3 main.py --display 
  

• Use DaSiamRPN method (based on Deep Learning):

  python3 main.py --display --method=DaSiamRPN
  

• Use custom data:

  python3 main.py --display --method=DaSiamRPN --video=sample1.mp4  --json=sample1.json --output=results.avi 
  

Using docker

Create docker image:

docker build -f Dockerfile -t obj_tracking:v01 .

Run this command to execute the program inside the container. Replace python3 main.py --help with your own command.

docker run -it --rm --net=host -e DISPLAY=$DISPLAY -w /app -v /tmp/.X11-unix/:/tmp/.X11-unix -v $HOME:/data obj_tracking:v01 python3 main.py --help

If you want to use the --display option make sure to run this command on the terminal:

xhost +