Method

Local-Global Motion Tracker [LGM]


Submitted on 19 Sep. 2021 12:33 by
Wang Gaoang (Zhejiang University)

Running time:0.08 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
The method tackles the association issue for long-
term tracking with the exclusive fully-exploited
motion information. We address the tracklet
embedding issue with the proposed reconstruct-to-
embed strategy based on deep graph convolutional
neural networks (GCN)
Parameters:
See manuscript
Latex Bibtex:
@inproceedings{wang2021track,
title={Track without Appearance: Learn Box and
Tracklet Embedding with Local and Global Motion
Patterns for Vehicle Tracking},
author={Wang, Gaoang and Gu, Renshu and Liu,
Zuozhu and Hu, Weijie and Song, Mingli and Hwang,
Jenq-Neng},
booktitle={Proceedings of the IEEE/CVF
International Conference on Computer Vision
(ICCV)},
pages={9876--9886},
year={2021}
}

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 73.14 % 74.61 % 72.31 % 80.53 % 82.16 % 76.38 % 84.74 % 85.85 %

Benchmark TP FP FN
CAR 32143 2249 1568

Benchmark MOTA MOTP MODA IDSW sMOTA
CAR 87.60 % 84.12 % 88.90 % 448 72.76 %

Benchmark MT rate PT rate ML rate FRAG
CAR 85.08 % 12.46 % 2.46 % 164

Benchmark # Dets # Tracks
CAR 33711 840

This table as LaTeX