Hasty implementation of a top view people tracker for the computer vision course in unitn.
- input video, video.mp4
- ouput video, result.mp4
- ground truth files
- estimated trajectories of selected pedestrians
- code (main.py, feature_swarm.py, util.py)
- report on this project
- numpy
- opencv
- the Rtree python package
In main.py you will obviousy find the main loop, feature_swarm contains the tracking algorithm and util is where I have stuffed all the utility stuff to avoid overly cluttering the code.
To run the project just run main.py with python3, no arguments needed; it will output a file named output.mp4.
This is not a long term project or anything, do not expect tests or whatever.
Note that the names of some parameters or functions might not always be the same in both the pseudocode (report) and the code.