A seminar presenting two thematically related journal papers:
“A Novel Performance Evaluation Methodology for Single-Target Trackers“.
(accepted to) IEEE TPAMI 2016 and “Robust scale-adaptive mean-shift for
tracking“. PRL 2014.
A novel performance evaluation methodology addresses the problem of
single-target tracker performance evaluation. We consider the performance
measures, the dataset and the evaluation system to be the most important
components of tracker evaluation and propose requirements for each of them. The
requirements are the basis of a new evaluation methodology that aims at a
and easily interpretable tracker comparison. The ranking-based methodology
addresses tracker equivalence in terms of statistical significance and
differences. A fully-annotated dataset with per-frame annotations with several
visual attributes is introduced. A multi-platform evaluation system allowing
easy integration of third-party trackers is presented as well. The proposed
evaluation methodology was used in a successful VOT challenges.
The mean-shift procedure was a popular object tracking algorithm since it was
fast, easy to implement and performs well in a range of conditions. We revived
this method and address the problem of scale adaptation and present a novel
theoretically justified scale estimation mechanism which relies solely on the
mean-shift procedure for the Hellinger distance. We also propose two
improvements of the mean-shift tracker that make the tracker more robust in the
presence of background clutter: (i) A novel histogram color weighting that
exploits the object neighbourhood to help discriminate the target. (ii) A
forward-backward consistency check and scale regularization. The performance is
demonstrated in the VOT2015 where it achieves a competitive results as modern
and more complex methods, yet running at 100fps on average.