Method

Stable at Any Speed: Speed-Driven Multi-Object Tracking with Learnable Kalman Filtering [SG-LKF]


Submitted on 23 May. 2025 15:37 by
Mengjun Chen (BIT)

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

Method Description:
/
Parameters:
/
Latex Bibtex:
@misc{gong2025stablespeedspeeddrivenmultiobject,
title={Stable at Any Speed: Speed-Driven
Multi-Object Tracking with Learnable Kalman
Filtering},
author={Yan Gong and Mengjun Chen and Hao
Liu and Gao Yongsheng and Lei Yang and Naibang
Wang and Ziying Song and Haoqun Ma},
year={2025},
eprint={2508.00358},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2508.00358},
}

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 90.75 % 85.84 % 91.08 % 88.48 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 92.96 % 99.07 % 95.92 % 36064 337 2730 3.03 % 44335 1235

Benchmark MT PT ML IDS FRAG
CAR 82.92 % 14.15 % 2.92 % 113 366

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