Our tracker is state-of-the-art on the MOT17 benchmark.

November 22, 2017


A Novel Multi-Detector Fusion Framework for Multi-Object Tracking


Authors: Roberto Henschel, Laura Leal-Taixé, Daniel Cremers, Bodo Rosenhahn

Abstract: In order to track all persons in a scene, the tracking-by-detection paradigm has proven to be a very effective approach. Yet, relying solely on a single detector is also a major limitation, as useful image information might be ignored. This work demonstrates how to incorporate several detectors into a tracking system, using a novel multi-object tracking formulation. We cast tracking as a weighted graph labeling problem, resulting in a binary quadratic program. As such problems are NP-hard, the solution can only be approximated. Based on the Frank-Wolfe algorithm, we present a new solver that is crucial to handle such difficult problems. As a result, the tracker can take information from many frames and different detectors holistically into account. When applied with head and full-body detections, the fusion helps to recover heavily occluded persons and to reduce false positives. Evaluation on pedestrian tracking is provided for multiple scenarios, showing superior results over single detector tracking and standard QP-solvers. Finally, our tracker performs state-of-the-art on the MOT16 benchmark and is the winner of the MOT17 challenge.

PDF:

MOTChallenge: MOT17 results

Results:

Drawing