Method

Observation-Centric SORT [on] [OC-SORT]
https://github.com/noahcao/OC_SORT

Submitted on 25 Mar. 2022 14:12 by
Travis Song (University of California, San Diego)

Running time:0.03 s
Environment:1 core @ 3.0 Ghz (Python)

Method Description:
Use the detections from PermanceTrack. OC-SORT uses Kalman
Filter for tracking without using appearance. It makes improvement
over the standard SORT by recognizing its limitation in occlusion and
non-linear motion. The inference speed of tracking given off-shelf
detection is 700fps on a i9@3GHz CPU. The method keeps online,
simple and realtime.
Parameters:
detection confidence threshold = 0.6,
IoU matching threshold = 0.3
min_hits = 3
Latex Bibtex:
@misc{cao2022observationcentric,
title={Observation-Centric SORT: Rethinking SORT for Robust
Multi-Object Tracking},
author={Jinkun Cao and Xinshuo Weng and Rawal Khirodkar
and Jiangmiao Pang and Kris Kitani},
url={https://arxiv.org/abs/2203.14360},
year={2022},
eprint={2203.14360},
archivePrefix={arXiv},
primaryClass={cs.CV}
}

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 90.64 % 85.71 % 91.30 % 88.41 %
PEDESTRIAN 64.01 % 74.73 % 64.71 % 92.21 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 93.37 % 98.93 % 96.07 % 36572 396 2597 3.56 % 46032 904
PEDESTRIAN 72.56 % 90.61 % 80.59 % 16956 1758 6412 15.80 % 22728 343

Benchmark MT PT ML IDS FRAG
CAR 81.23 % 15.85 % 2.92 % 225 471
PEDESTRIAN 44.67 % 35.74 % 19.59 % 161 813

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