Probability-Hypothesis-Density Filters for Pedestrian Tracking

:floppy_disk: code

Mobile Robotics Course - Winter 2021 - Professor Maani Ghaffari

We explore an object tracking filter that does not assume any association, unlike variants that the course discusses. The P.H.D. filter models the extent of track association as a Gaussian distribution. The project showed the success of the filter in certain settings and the general difficulty of estimating people (highly deformable objects) using low dimensional (bounding box) tracks.