Vehicle Trajctory Anomaly Detection

In this repository there is some of our work for a vision-based algorithm for vehicle trajectory anomaly detection.
This includes going off the road, stopping in the middle of the road, crossing a red light etc.
In this partial version, we rely on Object detection tools (such as detectron2) and trackers (such as Kalman fitler), in addition to a custom algorithm for recognizing trajectory anomalies.