Method

Continuous Energy Minimization [CEM]
http://research.milanton.net/contracking/

Submitted on 3 Jun. 2014 02:27 by
Anton Milan (University of Adelaide)

Running time:0.09 s
Environment:1 core @ >3.5 Ghz (Matlab + C/C++)

Method Description:
Multi-target tracking is performed as minimization
of a continuous energy that includes many important
aspects. Besides the image evidence, the energy
function takes into account physical constraints,
such as target dynamics, mutual exclusion, and track
persistence. A suitable optimization scheme that
alternates between continuous conjugate gradient
descent and discrete trans-dimensional jump moves is
able to find strong local minima of the proposed
non-convex energy.
Parameters:
Provided L-SVM detections with a score above 0.0
servce as input, along with the following
parameters:
wtEdyn=1
wtEexc=0.5
wtEper=0.5
wtEreg=0.25
lambda=0.15

The ooptimization is initialized with the Dynamic
Programming [DP] solution of [Pirsiavash et al.].
No appearance and no occlusion modeling is used in
this setting.
Latex Bibtex:
@article{Milan2014PAMI,
author = {Milan, A. and Roth, S. and Schindler, K.},
title = {Continuous Energy Minimization for Multitarget Tracking},
volume = {36},
issn = {0162-8828},
doi = {10.1109/TPAMI.2013.103},
number = {1},
journal = {IEEE TPAMI},
year = {2014},
pages = {58--72}
}

Detailed Results

From all 29 test sequences, our benchmark computes the commonly used tracking metrics CLEARMOT, MT/PT/ML, identity switches, and fragmentations [1,2]. The tables below show all of these metrics.


Benchmark MOTA MOTP MODA MODP
CAR 51.94 % 77.11 % 52.30 % 82.73 %
PEDESTRIAN 27.54 % 68.48 % 27.96 % 91.90 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 55.96 % 96.09 % 70.73 % 19819 807 15598 7.25 % 23188 935
PEDESTRIAN 36.73 % 80.82 % 50.51 % 8511 2020 14658 18.16 % 12687 520

Benchmark MT PT ML IDS FRAG
CAR 20.00 % 48.46 % 31.54 % 125 396
PEDESTRIAN 8.93 % 39.18 % 51.89 % 96 608

This table as LaTeX


[1] K. Bernardin, R. Stiefelhagen: Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. JIVP 2008.
[2] Y. Li, C. Huang, R. Nevatia: Learning to associate: HybridBoosted multi-target tracker for crowded scene. CVPR 2009.


eXTReMe Tracker