Method

SRK_ODESA(car) [on][at] [SRK_ODESA(mc)]


Submitted on 24 Dec. 2019 12:48 by
Viktor Porokhonskyy (Samsung R&D Institute Ukraine)

Running time:0.4 s
Environment:GPU (Python)

Method Description:
The corresponding solution employs tracking-by-detection approach. Its core is formed by an effective object embedding which enjoys several attractive properties. Namely, it is rather light-weight, offers expandability to the case of objects composed from multiple parts and demonstrates good generalization. The last property could be illustrated by the fact that no KITTI data was involved into the embedding training procedure.

The solution optimizes MOTA value.
Parameters:
private
Latex Bibtex:
@inproceedings{ODESA2020,
author = {Dmytro Mykheievskyi and
Dmytro Borysenko and
Viktor Porokhonskyy},
title = {Learning Local Feature Descriptors for Multiple Object Tracking},
booktitle = {ACCV},
year = {2020}
}

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 90.03 % 84.32 % 90.29 % 87.18 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 92.62 % 98.77 % 95.60 % 36258 451 2887 4.05 % 49307 2330

Benchmark MT PT ML IDS FRAG
CAR 82.62 % 15.08 % 2.31 % 90 501

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