An assessment of Multi Object Tracking on low framerate conditions

Anis Yassine Ben Mabrouk, Gabriele Facciolo, Rafael Grompone von Gioi, Axel Davy

Abstract:

The performance of state-of-the-art multiple object tracking methods is usually shown on high-framerate videos, for which objects
move very little between two consecutive frames. Nonetheless, there is an interest in working in the harder scenario of low framerate or similarly, strong motion. This article studies how much tracking performance is affected by a decrease in image frequency, and identifies the methods that are better suited for this alternative use-case.