Method

A Joint Multi-object Tracking Method with Center-based Feature Extraction and Occlusion Handling [CJMODT]
[Anonymous Submission]

Submitted on 15 Jul. 2022 13:55 by
[Anonymous Submission]

Running time:0.05 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
In this work, a joint multi-object tracking
algorithm is proposed. First, pairs of frames in
complicated environments are taken as input. A
center-based feature extraction framework is
designed for precisely detecting objects and
extracting their feature maps. Then, a ConvGRU
module is applied to learn permanent
representations by using historical spatio-
temporal information of objects.
Parameters:
Batchsize = 2
Epochs = 100
Iterations = 1600
Learning rate = 1.25*10^-4
Latex Bibtex:

Detailed Results

From all 29 test sequences, our benchmark computes the HOTA tracking metrics (HOTA, DetA, AssA, DetRe, DetPr, AssRe, AssPr, LocA) [1] as well as the CLEARMOT, MT/PT/ML, identity switches, and fragmentation [2,3] metrics. The tables below show all of these metrics.


Benchmark HOTA DetA AssA DetRe DetPr AssRe AssPr LocA
CAR 74.09 % 75.86 % 72.99 % 78.59 % 87.36 % 75.20 % 90.90 % 87.44 %
PEDESTRIAN 42.38 % 46.82 % 38.73 % 51.16 % 71.41 % 42.30 % 72.91 % 79.26 %

Benchmark TP FP FN
CAR 30683 3709 254
PEDESTRIAN 15116 8034 1468

Benchmark MOTA MOTP MODA IDSW sMOTA
CAR 87.44 % 86.08 % 88.48 % 357 75.02 %
PEDESTRIAN 56.55 % 75.48 % 58.95 % 557 40.54 %

Benchmark MT rate PT rate ML rate FRAG
CAR 75.69 % 20.46 % 3.85 % 335
PEDESTRIAN 36.77 % 42.27 % 20.96 % 724

Benchmark # Dets # Tracks
CAR 30937 991
PEDESTRIAN 16584 818

This table as LaTeX

This figure as: png pdf

This figure as: png pdf

[1] J. Luiten, A. Os̆ep, P. Dendorfer, P. Torr, A. Geiger, L. Leal-Taixé, B. Leibe: HOTA: A Higher Order Metric for Evaluating Multi-object Tracking. IJCV 2020.
[2] K. Bernardin, R. Stiefelhagen: Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. JIVP 2008.
[3] Y. Li, C. Huang, R. Nevatia: Learning to associate: HybridBoosted multi-target tracker for crowded scene. CVPR 2009.


eXTReMe Tracker