Modeling of Object Appearance by Normalized Adaptation for 3D Multiple Object Tracking

PETS09-S2L2 AVG-TownCentre

Short name:

MOANA

Benchmark:

Description:

(1) 3D Kalman-filter-based tracking for fully visible object(s)
(2) Cross-matching based on adaptive appearance modeling for partially occluded/grouped object(s)
(3) Re-identification based on adaptive appearance modeling for seriously occluded object(s)

Hardware:

4.20 GHz, 4 Cores

Detector:

Public

Processing:

Online

Last submitted:

December 06, 2017 (1 year ago)

Published:

September 04, 2018 at 01:28:28 CET

Submissions:

3

Open source:

No

Project page / code:

n/a

Reference:

Z. Tang, R. Gu, J. Hwang. Joint multi-view people tracking and pose estimation for 3D scene reconstruction. In Proc. IEEE Int. Conf. Multimedia Expo (ICME), 2018.

Benchmark performance:

MOTAMOTPFAFMTMLFPFNID Sw.FragSpecificationsDetector
52.756.32.528.4 % 22.0 % 2,2265,5511675864.20 GHz, 4 CoresPublic
IDF1ID PrecisionID Recall
62.470.156.2

Detailed performance:

Sequence MOTA IDF1 MOTP FAF GT MT ML FP FN ID Sw Frag
PETS09-S2L257.661.357.03.34240.5 % 7.1 % 1,4532,531107386
AVG-TownCentre46.164.055.11.722626.1 % 24.8 % 7733,02060200

Raw data:


MOANA