Near Online Multi-target Tracker with Aggregated Local Flow Descriptor with SDP detections

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

Short name:

NOMTwSDP16

Benchmark:

Description:

We trained a private detector (SDP + RPN) using the MOT16 training dataset. The tracker model/parameters are identical to NOMT_16. The detections for the test dataset is available at: http://www-personal.umich.edu/~wgchoi/MOT16-test-SDP.zip

The timing excludes detection time. With a K40 GPU, the detector runs at approximately 2 FPS.

@inproceedings{yang2016sdp,
title={Exploit All the Layers: Fast and Accurate CNN
Object Detector with Scale Dependent Pooling and
Cascaded Rejection Classifiers},
author={Fan Yang and Wongun Choi and Yuanqing Lin},
booktitle={Proceedings of the IEEE International
Conference on Computer Vision and Pattern Recognition},
year={2016}
}

Hardware:

2.4 GHz Xeon, 16 Cores

Detector:

Private

Processing:

Batch

Last submitted:

April 15, 2016 (3 years ago)

Published:

April 09, 2016 at 03:16:49 CET

Submissions:

2

Open source:

No

Project page / code:

n/a

Reference:

Benchmark performance:

MOTAMOTPFAFMTMLFPFNID Sw.FragSpecificationsDetector
62.279.60.932.5 % 31.1 % 5,11963,3524066422.4 GHz Xeon, 16 CoresPrivate
IDF1ID PrecisionID Recall
62.677.252.6

Detailed performance:

Sequence MOTA IDF1 MOTP FAF GT MT ML FP FN ID Sw Frag
MOT16-0154.155.975.90.42339.1 % 26.1 % 1982,7241216
MOT16-0372.768.980.60.714847.3 % 11.5 % 1,06127,319113175
MOT16-0661.366.978.60.622140.7 % 33.0 % 7263,6806488
MOT16-0749.950.679.41.55424.1 % 24.1 % 7527,3656175
MOT16-0842.940.080.01.16322.2 % 30.2 % 6758,8018391
MOT16-1250.360.380.00.68631.4 % 39.5 % 4973,6062227
MOT16-1439.853.373.01.616414.6 % 45.1 % 1,2109,85751170

Raw data:


NOMTwSDP16