Method

Combined Image- and World-Space Tracking in Traffic Scenes* [st] [on] [CIWT*]
https://github.com/aljosaosep/ciwt

Submitted on 12 Feb. 2017 20:16 by
Aljosa Osep (RWTH Aachen)

Running time:0.28 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
We use 3D object proposals to localize detections
accurately in 3D space and utilize 3D information
in our coupled 2D-3D filter, combined with
hypothesize-and-select framework.

* Using Regionlet detections
Parameters:
N/A
Latex Bibtex:
@inproceedings{Osep17ICRA,
title={Combined Image- and World-Space Tracking
in Traffic Scenes},
author={Osep, Aljosa and Mehner, Wolfgang
and Mathias, Markus and Leibe, Bastian},
booktitle={ICRA},
year={2017}
}

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 75.39 % 79.25 % 75.87 % 83.63 %
PEDESTRIAN 43.37 % 71.44 % 43.86 % 92.43 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 80.32 % 96.92 % 87.84 % 29985 954 7345 8.58 % 33605 1457
PEDESTRIAN 49.12 % 90.63 % 63.71 % 11408 1179 11818 10.60 % 13739 712

Benchmark MT PT ML IDS FRAG
CAR 49.85 % 39.85 % 10.31 % 165 660
PEDESTRIAN 13.75 % 51.55 % 34.71 % 112 901

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