A PDF "DM DE FREITAS - Robust object tracking using neural network based instance segmentation" provides a background on the project as well as the process for algorithm implementation.
DETECT contains a jupyter notebook used to perform instance segmentation on a video and save the results.
The various tracking algorithms developed are then provided i.e. Using the Gaussian Mixture Model and the Hungarian algorithm. Additional features are added in other models such as the amount of pixels an object occupies.
The requirements file is given and all the requirements must be installed before use.
Visualise then allows the results to be visualised on each frame and the video stitched back together.
An example output is given: https://www.youtube.com/watch?v=rGxS3IbGQwM&ab_channel=DanielDeFreitas