MOT17 Results

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


Showing only entries that use public detections!

TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
FWT
1. using public detections
6.9
51.3
±13.1
47.621.4% 35.2% 24,101247,9212,648 (47.2)4,279 (76.3)0.2Public
R. Henschel, L. Leal-Taixé, D. Cremers, B. Rosenhahn. A Novel Multi-Detector Fusion Framework for Multi-Object Tracking. In arXiv preprint arXiv:1705.08314, 2017.
jCC
2. using public detections
6.1
51.2
±14.5
54.520.9% 37.0% 25,937247,8221,802 (32.1)2,984 (53.2)1.8Public
M. Keuper, S. Tang, Y. Zhongjie, B. Andres, T. Brox, B. Schiele. A multi-cut formulation for joint segmentation and tracking of multiple objects. In arXiv preprint arXiv:1607.06317, 2016.
IMWIS
3. using public detections
7.6
51.0
±13.8
44.121.1% 36.2% 20,556252,9152,917 (52.9)3,128 (56.7)0.6Public
Anonymous submission
MHT_DAM
4. using public detections
7.8
50.7
±13.7
47.220.8% 36.9% 22,875252,8892,314 (41.9)2,865 (51.9)0.9Public
C. Kim, F. Li, A. Ciptadi, J. Rehg. Multiple Hypothesis Tracking Revisited. In ICCV, 2015.
HAF17
5. using public detections
7.3
50.6
±13.5
50.520.2% 36.8% 26,388250,0202,124 (38.1)2,938 (52.8)0.6Public
Anonymous submission
EDMT17
6. using public detections
7.1
50.0
±13.9
51.321.6% 36.3% 32,279247,2972,264 (40.3)3,260 (58.0)0.6Public
J. Chen, H. Sheng, Y. Zhang, Z. Xiong. Enhancing Detection Model for Multiple Hypothesis Tracking. In BMTT-PETS CVPRw, 2017.
DMAN
7. online method using public detections
8.2
48.2
±12.3
55.719.3% 38.3% 26,218263,6082,194 (41.2)5,378 (100.9)0.5Public
Anonymous submission
JCSTDDP
8. online method using public detections
10.3
47.6
±11.8
43.016.9% 34.2% 27,672264,4223,825 (72.0)8,688 (163.5)16.2Public
Anonymous submission
E2EM
9. online method using public detections
9.5
47.5
±14.5
48.816.5% 37.5% 20,655272,1873,632 (70.2)12,712 (245.6)29.6Public
Anonymous submission
CFWM
10. online method using public detections
9.5
47.0
±13.3
42.718.1% 36.3% 26,123268,1874,649 (88.6)7,015 (133.7)9.7Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
TD2L
11. using public detections
7.4
46.9
±13.7
51.921.9% 34.8% 43,892253,8402,006 (36.5)3,206 (58.3)25.4Public
Anonymous submission
PHD_DCM
12. online method using public detections
12.5
46.5
±13.8
47.616.9% 37.2% 23,859272,4305,649 (109.2)9,298 (179.8)1.6Public
Anonymous submission
MBKF
13. online method using public detections
11.6
45.7
±14.7
46.314.9% 41.8% 16,958286,2133,292 (66.8)7,659 (155.4)23.5Public
Anonymous submission
IOU17
14. using public detections
11.8
45.5
±13.6
39.415.7% 40.5% 19,993281,6435,988 (119.6)7,404 (147.8)1,522.9Public
E. Bochinski, V. Eiselein, T. Sikora. High-Speed Tracking-by-Detection Without Using Image Information. In International Workshop on Traffic and Street Surveillance for Safety and Security at IEEE AVSS 2017, 2017.
NN_LM
15. using public detections
11.7
45.1
±13.3
43.214.8% 46.2% 10,834296,4512,286 (48.2)2,463 (51.9)2.5Public
Anonymous submission
ESCNN
16. online method using public detections
12.2
44.9
±15.3
48.616.6% 35.8% 33,689269,8707,406 (142.0)14,489 (277.7)7.8Public
Anonymous submission
MEST
17. online method using public detections
12.3
44.7
±16.0
47.215.9% 34.9% 30,666275,8175,382 (105.3)9,210 (180.2)11.8Public
Anonymous submission
tmp_mot17
18. online method using public detections
13.6
43.8
±13.1
47.09.9% 44.7% 19,279295,0532,764 (57.9)10,459 (219.2)112.0Public
Anonymous submission
DP_NMS
19. using public detections
11.9
43.7
±14.0
36.912.6% 46.5% 10,048302,7284,942 (106.6)5,342 (115.3)137.7Public
Anonymous submission
PHD_CMAMM
20. online method using public detections
13.8
43.6
±14.1
43.914.9% 42.7% 26,988288,1902,908 (59.4)4,669 (95.4)3.3Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
SDF_IOU
21. online method using public detections
15.1
43.5
±13.8
35.015.6% 40.2% 25,612283,2379,909 (199.0)8,579 (172.3)71.0Public
Anonymous submission
TBD
22. using public detections
16.0
43.2
±14.7
36.316.3% 37.3% 45,131267,6417,570 (144.0)7,489 (142.5)1.5Public
Anonymous submission
DMPO
23. online method using public detections
12.6
43.1
±13.8
39.011.3% 46.4% 11,949304,6214,408 (95.8)6,679 (145.2)29.7Public
Anonymous submission
GM_PHD
24. online method using public detections
16.0
36.4
±14.1
33.94.1% 57.3% 23,723330,7674,607 (111.3)11,317 (273.5)38.4Public
V. Eiselein, D. Arp, M. Pätzold, T. Sikora. Real-time Multi-Human Tracking using a Probability Hypothesis Density Filter and multiple detectors. In 9th IEEE International Conference on Advanced Video and Signal-Based Surveillance, 2012.
GMPHD_KCF
25. online method using public detections
16.8
30.5
±13.7
35.79.6% 41.8% 107,802277,5426,774 (133.3)7,833 (154.2)3.3Public
T. Kutschbach, E. Bochinski, V. Eiselein, T. Sikora. Sequential Sensor Fusion Combining Probability Hypothesis Density and Kernelized Correlation Filters for Multi-Object Tracking in Video Data. In International Workshop on Traffic and Street Surveillance for Safety and Security at IEEE AVSS 2017, 2017.
SORTASMS
26. online method using public detections
19.7
25.4
±8.2
25.42.8% 49.6% 61,821353,0545,792 (154.8)10,096 (269.8)3.9Public
Anonymous submission
STSiam
27. online method using public detections
20.0
15.7
±19.4
19.52.6% 50.4% 93,155376,7945,730 (172.5)6,890 (207.4)0.8Public
Anonymous submission

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
21177572355564228

Difficulty Analysis

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

MOT17-03-SDP

MOT17-03-SDP

(67.8% MOTA)

MOT17-03-FRCNN

MOT17-03-FRCNN

(54.7% MOTA)

MOT17-06-SDP

MOT17-06-SDP

(46.8% MOTA)

...

...

MOT17-14-DPM

MOT17-14-DPM

(17.1% MOTA)

MOT17-14-FRCNN

MOT17-14-FRCNN

(16.5% MOTA)


Evaluation Measures

Lower is better. Higher is better.
Measure Better Perfect Description
Avg Rank lower 1 This is the rank of each tracker averaged over all present evaluation measures.
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.

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.
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.