DD_TAMA19: Online Multiple Pedestrian Tracking with Deep Temporal Appearance Matching Association


Video not available.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Short name:

DD_TAMA19

Detector:

Public

Description:

In online multi-target tracking, it is of great importance to model appearance and geometric similarity between pedestrians which have been tracked and appeared in a new frame. The dimension of the inherent feature vector in the appearance model is higher than that in the geometric model, which causes many problems in general. However, the recent success of deep learning-based methods makes it possible to handle high dimensional appearance information successfully. Among many deep networks, the Siamese network with triplet loss is popularly
adopted as an appearance feature extractor. Since the Siamese network can extract features of each input independently, it is possible to update and maintain target-specific features. However, it is not suitable for multi-target settings that require comparison with other inputs. In this paper, to address this issue, we propose a novel track appearance model based on the joint-inference network. The proposed method enables a comparison of two inputs to be used for adaptive appearance modeling, and contributes to disambiguating the process of target-observation matching and consolidating identity consistency. Diverse experimental results support the effectiveness of our method. Our work has been awarded as a 3rd-highest tracker on MOTChallenge19, held in CVPR2019.1 The code is available on https://github.com/yyc9268/Deep-TAMA.

Reference:

Y. Yoon, D. Kim, Y. Song, K. Yoon, M. Jeon. Online Multiple Pedestrians Tracking using Deep Temporal Appearance Matching Association. In Information Sciences, 2020.

Last submitted:

June 13, 2019 (5 years ago)

Published:

June 16, 2019 at 00:00:00 CET

Submissions:

1

Open source:

No

Hardware:

3.0GHZ, 1 Core, TITAN X

Runtime:

0.2 Hz

Benchmark performance:

Sequence MOTA IDF1 HOTA MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
CVPR 2019 Tracking Challenge47.648.70.0342 (27.2)297 (23.6)38,194252,93454.888.90.00.00.00.00.00.00.08.52,437 (44.4)3,887 (70.9)

Detailed performance:

Sequence MOTA IDF1 HOTA MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
CVPR19-0464.059.20.0268908,225105,49966.996.30.00.00.00.00.00.00.04.09001,325
CVPR19-0626.132.80.02610015,27980,83038.476.70.00.00.00.00.00.00.015.28141,420
CVPR19-0753.949.60.038151,86513,16160.291.40.00.00.00.00.00.00.03.2237288
CVPR19-0813.826.00.0109212,82553,44430.965.10.00.00.00.00.00.00.015.9486854

Raw data: