Learning Data Association for Multi-Object Tracking using Only Coordinates
This work addresses the challenge of tracking multiple objects in videos, particularly in crowded or dynamic scenes, by introducing a novel method that avoids reliance on traditional metrics like intersection over union or motion priors, representing an incremental improvement over existing trackers.
The paper tackles the data association problem in multi-object tracking by proposing a Transformer-based module that uses only bounding box coordinates to estimate affinity scores between tracks, achieving state-of-the-art performance on DanceTrack and KITTIMOT datasets and competitive results on MOT17.
We propose a novel Transformer-based module to address the data association problem for multi-object tracking. From detections obtained by a pretrained detector, this module uses only coordinates from bounding boxes to estimate an affinity score between pairs of tracks extracted from two distinct temporal windows. This module, named TWiX, is trained on sets of tracks with the objective of discriminating pairs of tracks coming from the same object from those which are not. Our module does not use the intersection over union measure, nor does it requires any motion priors or any camera motion compensation technique. By inserting TWiX within an online cascade matching pipeline, our tracker C-TWiX achieves state-of-the-art performance on the DanceTrack and KITTIMOT datasets, and gets competitive results on the MOT17 dataset. The code will be made available upon publication.