A multi target tracker based on Gaussian Mixture Probability Density Hypothesis Density Filter using Determinantal Point Processes pruning.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
- CMake 2.8 or newer
- OpenCV 3.4 (DNN library is required)
- Eigen 3.2
Download the project:
git clone https://github.com/fjorquerauribe/multitarget-tracking.git
Build the project:
cd multitarget-tracking
mkdir build && cd build
cmake ..
make
Download the MOT Challenge datasets https://motchallenge.net/
Create a symbolic link to the MOT Challenge datasets folder:
ln -s path/to/datasets/ data
Copy script to run Gaussian Mixture PHD filter:
cp ../scripts/start_gm_phd.sh .
Example: run Gaussian Mixture PHD filter over MOT16-02 sequence with public detections:
./start_gm_phd.sh MOT16 train MOT16-02 public 1
Example 2: run PHD filter and YOLO detector over MOT16-02 sequence (we assume YOLO .cfg, .weights and .names files are into data/yolo folder)
./start_phd.sh MOT16 train MOT16-02 100 yolo 0.5
Demo video https://www.youtube.com/watch?v=yVi9aQtO6fU
- Felipe Jorquera Uribe - fjorquerauribe
- Sergio Hernández - shernandez
See also the list of contributors who participated in this project.
This project is licensed under the Apache 2 license - see the LICENSE.txt file for details.