Online Multi-Object Tracking with Hierarchical Data Association using GMPHD filter

MOT16-01 MOT16-03 MOT16-06 MOT16-07 MOT16-08 MOT16-12 MOT16-14

Short name:

GMPHD_HDA

Benchmark:

Description:

Stage 1: low-level association between targets states and observations(i.e., detection responses), frame by frame
-We only used detection reponses with the positive confidence value among provided detection results

Stage 2: short tracklet elimination under the set length(parameter)
-Minimum tracklet length for mid-level association : 10

Stage 3: mid-level association between rough tracklets(almost fragmented or ID swtiched)
-Only Motion and size are used
-Appearance is not used

Hardware:

3.5 GHz, 4 Cores

Detector:

Public

Processing:

Online

Last submitted:

August 07, 2016 (11 months ago)

Published:

August 09, 2016 at 22:37:16 CET

Submissions:

1

Open source:

No

Project page / code:

n/a

Reference:

Y. Song, M. Jeon. Online Multiple Object Tracking with the Hierarchically Adopted GM-PHD Filter using Motion and Appearance. In IEEE/IEIE The International Conference on Consumer Electronics (ICCE) Asia, 2016.

Benchmark performance:

MOTAMOTPFAFMTMLFPFNID Sw.FragSpecificationsDetector
30.575.40.94.6 % 59.7 % 5,169120,9705397313.5 GHz, 4 CoresPublic

Detailed performance:

Sequence MOTA MOTP FAF GT MT ML FP FN ID Sw Frag
MOT16-0123.972.60.3238.7 % 56.5 % 1514,6952135
MOT16-0337.475.71.11484.7 % 40.5 % 1,66763,599206293
MOT16-0628.772.81.22218.1 % 59.3 % 1,4306,71776116
MOT16-0725.174.30.6541.9 % 64.8 % 28011,9005191
MOT16-0820.078.51.0630.0 % 57.1 % 59912,7236862
MOT16-1221.875.80.7862.3 % 60.5 % 6315,7906242
MOT16-1413.474.20.51643.0 % 76.8 % 41115,5465592

Raw data:


GMPHD_HDA