Full Text:   <54>

CLC number: 

On-line Access: 2020-11-16

Received: 2020-06-04

Revision Accepted: 2020-10-08

Crosschecked: 0000-00-00

Cited: 0

Clicked: 131

Citations:  Bibtex RefMan EndNote GB/T7714

-   Go to

Article info.
Open peer comments

Journal of Zhejiang University SCIENCE C 1998 Vol.-1 No.-1 P.

http://doi.org/10.1631/FITEE.2000272


Learning associative affinity network for multiple object tracking


Author(s):  Liang MA, Qiaoyong ZHONG, Yingying ZHANG, Di XIE, Shiliang PU

Affiliation(s):  Hikvision Research Institute, Hangzhou 310000, China

Corresponding email(s):   maliang6@hikvision.com, zhongqiaoyong@hikvision.com, zhangyingying7@hikvision.com, xiedi@hikvision.com, pushiliang.hri@hikvision.com

Key Words:  Multiple object tracking, Deep neural network, Affinity learning


Liang MA, Qiaoyong ZHONG , Yingying ZHANG , Di XIE , Shiliang PU. Learning associative affinity network for multiple object tracking[J]. Frontiers of Information Technology & Electronic Engineering, 1998, -1(-1): .

@article{title="Learning associative affinity network for multiple object tracking",
author="Liang MA, Qiaoyong ZHONG , Yingying ZHANG , Di XIE , Shiliang PU",
journal="Frontiers of Information Technology & Electronic Engineering",
volume="-1",
number="-1",
pages="",
year="1998",
publisher="Zhejiang University Press & Springer",
doi="10.1631/FITEE.2000272"
}

%0 Journal Article
%T Learning associative affinity network for multiple object tracking
%A Liang MA
%A Qiaoyong ZHONG
%A Yingying ZHANG
%A Di XIE
%A Shiliang PU
%J Journal of Zhejiang University SCIENCE C
%V -1
%N -1
%P
%@ 2095-9184
%D 1998
%I Zhejiang University Press & Springer
%DOI 10.1631/FITEE.2000272

TY - JOUR
T1 - Learning associative affinity network for multiple object tracking
A1 - Liang MA
A1 - Qiaoyong ZHONG
A1 - Yingying ZHANG
A1 - Di XIE
A1 - Shiliang PU
J0 - Journal of Zhejiang University Science C
VL - -1
IS - -1
SP -
EP -
%@ 2095-9184
Y1 - 1998
PB - Zhejiang University Press & Springer
ER -
DOI - 10.1631/FITEE.2000272


Abstract: 
We propose a joint feature and metric learning deep neural network architecture, called the associative affinity network (AAN), as an affinity model for multiple object tracking (MOT) in videos. The AAN learns the associative affinity between tracklets and detections across frames in an end-to-end manner. Considering flawed detections, the AAN jointly learns bounding box regression, classification and affinity regression via proposed multi-task loss. Contrary to networks that are trained with ranking loss, we directly train a binary classifier to learn the associative affinity of each tracklet-detection pair and use a matching cardinality loss to capture information among candidate pairs. The AAN learns a discriminative affinity model for data association to tackle MOT, and can also perform single object tracking. Based on the AAN, we propose a simple multi-object tracker that achieves competitive performance on the public MOT16 and MOT17 test datasets.

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Open peer comments: Debate/Discuss/Question/Opinion

<1>

Please provide your name, email address and a comment





Journal of Zhejiang University-SCIENCE, 38 Zheda Road, Hangzhou 310027, China
Tel: +86-571-87952783; E-mail: cjzhang@zju.edu.cn
Copyright © 2000 - Journal of Zhejiang University-SCIENCE