1.WHAT IS THE GOAL OF MOT ?
Estimating the trajectory of objects in a video clip is the goal of Multiple Object Tracking (MOT). Certain MOT standards only consider tracking objects from a few defined categories, such as: Pedestrians and vehicles, the actual number of types of interest is enormous.
2.WHEN TETA USING ITS PERFORMANCE?
The proposed TETA metric separates classification and tracking performance. To evaluate the performance of localization and attribution, we group the prediction results within each cluster using the ground truth bounding box of each target class as an anchor for each cluster.
3.HOW THEY EXTENDED TRACKING?
To extend tracking to wider scenarios, they propose to separate classification and tracking during model design and evaluation of multi-category MOT. To do this, they created a new model called Track Every Thing Tracker and a new metric called Track Every Thing Accuracy (TETA)