3D MOT 2015 Results

Click on a measure to sort the table accordingly. See below for a more detailed description.


Showing only entries that use public detections!


Benchmark Statistics

TrackerMOTAIDF1MOTPMTMLFPFNRecallPrecisionFAFID Sw. FragHz
GPDBN
1. online method using public detections
49.8
±0.0
0.0
±0.0
62.2 69 (25.7)46 (17.2)1,813 6,300 62.5 85.3 2.0 311 (5.0)386 (6.2)0.1
T. Klinger, F. Rottensteiner, C. Heipke. Probabilistic multi-person localisation and tracking in image sequences. In ISPRS Journal of Photogrammetry and Remote Sensing, 2017.
DBN
2. online method using public detections
51.1
±0.0
0.0
±0.0
61.0 77 (28.7)48 (17.9)2,077 5,746 65.8 84.2 2.3 380 (5.8)418 (6.4)0.1
T. Klinger, F. Rottensteiner, C. Heipke. Probabilistic Multi-Person Tracking using Dynamic Bayes Networks. In ISPRS Workshop on Image Sequence Analysis (ISA), 2015.
MCFPHD
3. using public detections
39.9
±0.0
0.0
±0.0
53.6 69 (25.7)45 (16.8)3,029 6,700 60.1 76.9 3.4 363 (6.0)529 (8.8)17.7
N. Wojke, D. Paulus. Global data association for the Probability Hypothesis Density filter using network flows. In 2016 IEEE International Conference on Robotics and Automation, ICRA, 2016.
MPLT
4. online method using public detections
54.2
±0.0
48.8
±0.0
60.7 82 (30.6)56 (20.9)2,385 4,930 70.6 83.3 2.7 366 (5.2)538 (7.6)0.4
F. Fan Yang, S. Nakamura. Using panoramic videos for multi-person localization and tracking in a 3D panoramic coordinate. In ICASSP, 2020.
MOANA
5. online method using public detections
52.7
±0.0
62.4
±0.0
56.3 76 (28.4)59 (22.0)2,226 5,551 66.9 83.5 2.5 167 (2.5)586 (8.8)19.4
Z. Tang, J. Hwang. MOANA: An online learned adaptive appearance model for robust multiple object tracking in 3D. In IEEE Access, 2019.
LPSFM
6. using public detections
35.9
±49.4
0.0
±0.0
54.0 37 (13.8)58 (21.6)2,031 8,206 51.1 80.9 2.3 520 (10.2)601 (11.8)inf
L. Leal-Taixé, G. Pons-Moll, B. Rosenhahn. Everybody needs somebody: modeling social and grouping behavior on a linear programming multiple people tracker. In IEEE International Conference on Computer Vision Workshops (ICCVW). 1st Workshop on Modeling, Simulation and Visual Analysis of Large Crowds, 2011.
SVT
7. using public detections
34.2
±0.0
0.0
±0.0
55.8 30 (11.2)68 (25.4)3,057 7,454 55.6 75.3 3.5 532 (9.6)611 (11.0)1.9
Longyin Wen, Zhen Lei, Ming-Ching Chang, Honggang Qi, Siwei Lyu. Multi-Camera Multi-Target Tracking with Space-Time-View Hyper-graph. IJCV, 2016.
LP3D
8. using public detections
35.9
±49.4
0.0
±0.0
53.3 56 (20.9)44 (16.4)3,588 6,593 60.7 74.0 4.0 580 (9.6)659 (10.9)inf
MOT baseline: Linear programming on 3D image coordinates.
AMIR3D
9. online method using public detections
25.0
±9.5
0.0
±0.0
55.6 8 (3.0)74 (27.6)2,038 9,084 45.9 79.1 2.3 1,462 (31.9)1,647 (35.9)1.2
A. Sadeghian, A. Alahi, S. Savarese. Tracking The Untrackable: Learning To Track Multiple Cues with Long-Term Dependencies. In ICCV, 2017.
KalmanSFM
10. online method using public detections
25.0
±8.5
0.0
±0.0
53.6 18 (6.7)39 (14.6)3,161 7,599 54.7 74.4 3.6 1,838 (33.6)1,686 (30.8)30.6
S. Pellegrini, A. Ess, K. Schindler, L. Gool. You'll never walk alone: Modeling social behavior for multi-target tracking.. In ICCV, 2009.
SequencesFramesTrajectoriesBoxes
288626816789

Difficulty Analysis

Sequence difficulty (from easiest to hardest, measured by average MOTA)

PETS09-S2L2

PETS09-S2L2

(33.0 MOTA)

AVG-TownCentre

AVG-TownCentre

(15.0 MOTA)


Evaluation Measures

Lower is better. Higher is better.
Measure Better Perfect Description
MOTA higher 100 % Multiple Object Tracking Accuracy [1]. This measure combines three error sources: false positives, missed targets and identity switches.
MOTP higher 100 % Multiple Object Tracking Precision [1]. The misalignment between the annotated and the predicted bounding boxes.
IDF1 higher 100 % ID F1 Score [2]. The ratio of correctly identified detections over the average number of ground-truth and computed detections.
FAF lower 0 The average number of false alarms per frame.
MT higher 100 % Mostly tracked targets. The ratio of ground-truth trajectories that are covered by a track hypothesis for at least 80% of their respective life span.
ML lower 0 % Mostly lost targets. The ratio of ground-truth trajectories that are covered by a track hypothesis for at most 20% of their respective life span.
FP lower 0 The total number of false positives.
FN lower 0 The total number of false negatives (missed targets).
ID Sw. lower 0 The total number of identity switches. Please note that we follow the stricter definition of identity switches as described in [3].
Frag lower 0 The total number of times a trajectory is fragmented (i.e. interrupted during tracking).
Hz higher Inf. Processing speed (in frames per second excluding the detector) on the benchmark. The frequency is provided by the authors and not officially evaluated by the MOTChallenge.

Legend

Symbol Description
online method This is an online (causal) method, i.e. the solution is immediately available with each incoming frame and cannot be changed at any later time.
using public detections This method used the provided detection set as input.
using public detections This method used the provided detection set as input.
new This entry has been submitted or updated less than a week ago.
new This entry has been submitted or updated less than a week ago.

References:


[1] Bernardin, K. & Stiefelhagen, R. Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. Image and Video Processing, 2008(1):1-10, 2008.
[2] Ristani, E., Solera, F., Zou, R., Cucchiara, R. & Tomasi, C. Performance Measures and a Data Set for Multi-Target, Multi-Camera Tracking. In ECCV workshop on Benchmarking Multi-Target Tracking, 2016.
[3] Li, Y., Huang, C. & Nevatia, R. Learning to associate: HybridBoosted multi-target tracker for crowded scene. In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2009.