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
yt_face
1. online method using public detections new
9.9
52.6
±13.1
51.523.0% 35.9% 23,894241,4892,047 (35.8)2,827 (49.4)2.2Public
Anonymous submission
SST
2. online method using public detections
16.1
52.4
±15.9
49.521.4% 30.7% 25,423234,5928,431 (144.3)14,797 (253.3)6.3Public
ShiJie Sun, Naveed Akhtar, Ajmal Mian (UWA and ShiYuan Research Institute)
CNN_search
3. using public detections
8.5
52.2
±13.8
50.721.4% 33.3% 23,565243,2323,059 (53.8)4,702 (82.6)3.7Public
Liang Ma, Yingying Zhang, Qiaoyong Zhong, Di Xie and Shiliang Pu (Hikvision Research Institute)
eHAF17
4. using public detections
10.1
51.8
±13.2
54.723.4% 37.9% 33,212236,7721,834 (31.6)2,739 (47.2)0.7Public
TCSVT-02141-2018
AFN17
5. using public detections
10.3
51.5
±13.0
46.920.6% 35.5% 22,391248,4202,593 (46.3)4,308 (77.0)1.8Public
Anonymous submission
FWT
6. using public detections
12.7
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. Fusion of Head and Full-Body Detectors for Multi-Object Tracking. In Trajnet CVPRW, 2018.
jCC
7. using public detections
10.8
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
8. using public detections
12.9
51.0
±13.8
44.121.1% 36.2% 20,556252,9152,917 (52.9)3,128 (56.7)0.6Public
Anonymous submission
MOTDT17
9. online method using public detections
11.6
50.9
±11.9
52.717.5% 35.7% 24,069250,7682,474 (44.5)5,317 (95.7)18.3Public
Anonymous ICME submission
MHT_DAM
10. using public detections
13.2
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.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
HAF17
11. using public detections
12.9
50.6
±13.5
50.520.2% 36.8% 26,388250,0202,124 (38.1)2,938 (52.8)0.6Public
Anonymous submission
EDMT17
12. using public detections
12.2
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.
IDGA
13. using public detections
11.1
49.9
±12.2
50.322.1% 36.7% 37,060243,1482,426 (42.6)3,846 (67.6)59.2Public
Anonymous submission
TEM
14. using public detections
16.7
49.1
±12.6
45.417.0% 38.3% 22,119261,7973,439 (64.2)3,881 (72.4)8.2Public
Anonymous submission
DMAN
15. online method using public detections
14.8
48.2
±12.3
55.719.3% 38.3% 26,218263,6082,194 (41.2)5,378 (100.9)0.5Public
Anonymous submission
PHD_GSDL17
16. online method using public detections
16.3
48.0
±13.6
49.617.1% 35.6% 23,199265,9543,998 (75.6)8,886 (168.1)6.7Public
Z. Fu, P. Feng, F. Angelini, J. Chambers, S. Naqvi. Particle PHD Filter based Multiple Human Tracking using Online Group-Structured Dictionary Learning. In IEEE Access, 2018.
FLT
17. online method using public detections
16.2
47.7
±13.0
48.916.5% 37.7% 18,991272,8423,534 (68.4)12,635 (244.7)24.9Public
Anonymous submission
E2EM
18. online method using public detections
15.8
47.5
±14.5
48.816.5% 37.5% 20,655272,1873,632 (70.2)12,712 (245.6)29.6Public
Anonymous submission
CFWM
19. online method using public detections
16.5
47.0
±13.3
42.718.1% 36.3% 26,123268,1874,649 (88.6)7,015 (133.7)9.7Public
Anonymous submission
TD2L
20. using public detections
12.5
46.9
±13.7
51.921.9% 34.8% 43,892253,8402,006 (36.5)3,206 (58.3)25.4Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
SNM17
21. online method using public detections
23.3
46.8
±13.8
43.416.2% 37.1% 25,104271,0424,213 (81.1)9,891 (190.3)0.8Public
Anonymous submission
PHD_DCM
22. online method using public detections
21.6
46.5
±13.8
47.616.9% 37.2% 23,859272,4305,649 (109.2)9,298 (179.8)1.6Public
Anonymous submission
PHD_PM_OM
23. online method using public detections
21.9
46.2
±13.5
44.316.3% 40.0% 23,976276,3013,237 (63.4)5,218 (102.3)0.6Public
Anonymous submission
COMOT
24. online method using public detections new
16.3
46.1
±13.4
47.913.6% 42.5% 20,775281,1532,019 (40.2)4,480 (89.3)5.0Public
Anonymous submission
ts_WCFMT
25. online method using public detections
20.3
46.1
±13.3
48.517.7% 33.9% 28,607271,5613,824 (73.7)6,891 (132.9)1.0Public
Anonymous submission
IDOHMPT
26. online method using public detections
21.2
46.0
±13.1
44.116.8% 36.6% 30,873268,2215,768 (109.9)9,663 (184.2)8.1Public
Anonymous submission
MTS_CNN
27. online method using public detections
21.8
45.5
±12.7
41.417.3% 36.9% 33,774269,4934,033 (77.2)6,643 (127.2)2.8Public
Anonymous submission
IOU17
28. using public detections
21.4
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.
LM_NN
29. using public detections
19.8
45.1
±13.3
43.214.8% 46.2% 10,834296,4512,286 (48.2)2,463 (51.9)2.5Public
Anonymous submission
FPSN
30. online method using public detections
21.0
44.9
±13.9
48.416.5% 35.8% 33,757269,9527,136 (136.8)14,491 (277.8)10.1Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
MTS_CNN_ML
31. online method using public detections
25.2
44.8
±12.7
37.717.1% 37.6% 30,661276,8363,798 (74.6)7,085 (139.1)0.4Public
Anonymous submission
reID2track
32. online method using public detections
25.7
44.6
±14.3
39.915.8% 39.7% 22,451284,2136,134 (123.6)13,786 (277.8)9.0Public
Anonymous submission
Q_ls
33. online method using public detections
26.9
44.0
±13.7
40.115.7% 39.0% 41,442269,1495,136 (98.2)6,519 (124.7)0.6Public
Anonymous submission
ERC
34. online method using public detections
21.0
43.8
±13.1
49.213.8% 44.5% 27,729286,9172,299 (46.8)5,671 (115.4)5.9Public
Anonymous submission
PeriodMOT
35. online method using public detections
23.3
43.8
±13.2
40.914.7% 42.0% 21,941290,1944,910 (101.1)6,649 (136.9)66.9Public
Anonymous submission
tmp_mot17
36. online method using public detections
23.5
43.8
±13.1
47.09.9% 44.7% 19,279295,0532,764 (57.9)10,459 (219.2)112.0Public
Anonymous submission
GMPHD_SHA
37. online method using public detections
24.2
43.7
±12.5
39.211.7% 43.0% 25,935287,7583,838 (78.3)5,056 (103.2)9.2Public
Anonymous submission
PHD_CMAMM
38. online method using public detections
24.7
43.6
±14.1
43.914.9% 42.7% 26,988288,1902,908 (59.4)4,669 (95.4)1.1Public
Anonymous submission
SDF_IOU
39. online method using public detections
26.5
43.5
±13.8
35.015.6% 40.2% 25,612283,2379,909 (199.0)8,579 (172.3)71.0Public
Anonymous submission
EAMTT
40. online method using public detections
25.8
42.6
±13.3
41.812.7% 42.7% 30,711288,4744,488 (91.8)5,720 (117.0)1.4Public
R. Sanchez-Matilla, F. Poiesi, A. Cavallaro. Online Multi-target Tracking with Strong and Weak Detections. In Computer Vision -- ECCV 2016 Workshops, 2016.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
ATIOU
41. online method using public detections
23.0
42.4
±15.0
40.312.9% 44.9% 12,720307,5804,804 (105.6)12,521 (275.3)123.3Public
Anonymous submission
MPTV1
42. online method using public detections
29.4
41.5
±13.0
40.211.9% 43.9% 21,853302,6515,418 (116.9)13,682 (295.1)0.5Public
Anonymous submission
GMPHD_KCF
43. online method using public detections
29.3
39.6
±13.6
36.68.8% 43.3% 50,903284,2285,811 (117.1)7,414 (149.4)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.
GM_PHD
44. online method using public detections
27.4
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.

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
21177572355564228

Difficulty Analysis

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

MOT17-03-SDP

MOT17-03-SDP

(71.0% MOTA)

MOT17-03-FRCNN

MOT17-03-FRCNN

(57.2% MOTA)

MOT17-06-SDP

MOT17-06-SDP

(51.4% MOTA)

...

...

MOT17-14-DPM

MOT17-14-DPM

(19.2% MOTA)

MOT17-14-FRCNN

MOT17-14-FRCNN

(19.1% 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.