Method

AGTrack: Adaptive Kalman Filtering and Geometric Similarity for 3D MOT [AGTrack]
[Anonymous Submission]

Submitted on 26 Jan. 2026 11:39 by
[Anonymous Submission]

Running time:0.05 s
Environment:8 cores @ >3.5 Ghz (Python)

Method Description:
AGTrack is a novel 3D multi-object tracking
framework designed for autonomous driving
perception systems. It introduces a geometry-aware
3D Rotation-Shape IoU (3D-RSIoU) metric that
comprehensively models rotation alignment and
shape characteristics for more accurate data
association. Additionally, the framework
incorporates an adaptive Kalman filter that
dynamically adjusts noise parameters based on
real-time detection quality and prediction-
observation discrepancies, enhancing state
estimation robustness in complex motion scenarios.
The system operates within a single-stage matching
architecture, maintaining high efficiency while
achieving accuracy comparable to multi-stage
methods.
Parameters:
\lamda_1=1, \lamda_2=1, \lamda_3=0.5, lamda_4=0.5

Latex Bibtex:

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 91.75 % 86.90 % 91.83 % 89.57 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 94.21 % 98.62 % 96.36 % 37216 521 2288 4.68 % 42902 677

Benchmark MT PT ML IDS FRAG
CAR 87.23 % 4.77 % 8.00 % 30 54

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