Discrete-Continuous Energy Minimization with Exclusion

TUD-Crossing PETS09-S2L2 ETH-Jelmoli ETH-Linthescher ETH-Crossing AVG-TownCentre ADL-Rundle-1 ADL-Rundle-3 KITTI-16 KITTI-19

Short name:

DCO_X

Benchmark:

Description:

We formulate multi-target tracking as a discrete-continuous optimization problem that handles both data association and trajectory estimation in its respective natural domain and allows leveraging powerful methods for multi-model fitting. Data association is performed using discrete optimization with label costs, yielding near optimality. Trajectory estimation is posed as a continuous fitting problem which is used in turn to update the label costs.

Hardware:

2.7 GHz, 1 Core

Detector:

Public

Processing:

Batch

Last submitted:

June 02, 2015 (2 years ago)

Published:

July 28, 2015 at 17:01:02 CET

Submissions:

2

Open source:

Yes

Project page / code:

Reference:

A. Milan, K. Schindler, S. Roth. Multi-Target Tracking by Discrete-Continuous Energy Minimization. In IEEE PAMI, 2016.

Benchmark performance:

MOTAMOTPFAFMTMLFPFNID Sw.FragSpecificationsDetector
19.671.41.85.1 % 54.9 % 10,65238,2325218192.7 GHz, 1 CorePublic
IDF1ID PrecisionID Recall
31.544.324.4

Detailed performance:

Sequence MOTA IDF1 MOTP FAF GT MT ML FP FN ID Sw Frag
TUD-Crossing57.355.273.70.21315.4 % 15.4 % 404181317
PETS09-S2L237.527.370.71.5424.8 % 16.7 % 6385,200189209
ETH-Jelmoli30.248.871.61.24515.6 % 35.6 % 5131,2382059
ETH-Linthescher17.023.974.90.21974.1 % 76.6 % 2117,1792432
ETH-Crossing16.528.375.30.2263.8 % 69.2 % 4778747
AVG-TownCentre8.223.269.91.72262.7 % 69.5 % 7635,7663084
ADL-Rundle-110.034.072.06.83215.6 % 28.1 % 3,3904,92068111
ADL-Rundle-316.931.672.64.2449.1 % 25.0 % 2,6475,7306974
KITTI-1634.041.273.80.7170.0 % 11.8 % 1459582024
KITTI-1917.439.365.41.2623.2 % 29.0 % 1,3103,05547149
Venice-113.126.571.72.1170.0 % 29.4 % 9482,9813753

Raw data:


DCO_X