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
HIK_MOT17
1. using public detections new
10.2
53.9
±13.7
54.323.7% 32.0% 27,656230,0422,386 (40.3)4,192 (70.8)5.4Public
NOTBD
2. using public detections
15.6
53.9
±12.7
51.221.5% 35.6% 28,912228,3562,964 (49.8)3,600 (60.5)0.3Public
Anonymous submission
yt_face
3. online method using public detections
12.9
52.6
±13.1
51.523.0% 35.9% 23,894241,4892,047 (35.8)2,827 (49.4)2.2Public
Anonymous submission
SST
4. online method using public detections
19.3
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
5. using public detections
11.4
52.2
±13.8
50.721.4% 33.3% 23,565243,2323,059 (53.8)4,702 (82.6)3.7Public
Anonymous submission
eHAF17
6. using public detections
12.7
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
7. using public detections
13.1
51.5
±13.0
46.920.6% 35.5% 22,391248,4202,593 (46.3)4,308 (77.0)1.8Public
Anonymous submission
FWT
8. using public detections
15.3
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
9. using public detections
13.8
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
10. using public detections
16.1
51.0
±13.8
44.121.1% 36.2% 20,556252,9152,917 (52.9)3,128 (56.7)0.6Public
TCSVT-02160-2018
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
MOTDT17
11. online method using public detections
14.3
50.9
±11.9
52.717.5% 35.7% 24,069250,7682,474 (44.5)5,317 (95.7)18.3Public
C. Long, A. Haizhou, Z. Zijie, S. Chong. Real-time Multiple People Tracking with Deeply Learned Candidate Selection and Person Re-identification. In ICME, 2018.
MHT_DAM
12. using public detections
16.4
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.
ORCtracker
13. online method using public detections new
19.5
50.7
±13.7
43.117.0% 35.2% 20,440249,7918,069 (144.8)11,188 (200.8)3,760.7Public
Anonymous submission
HAF17
14. using public detections
16.1
50.6
±13.5
50.520.2% 36.8% 26,388250,0202,124 (38.1)2,938 (52.8)0.6Public
Anonymous submission
TLMHT
15. using public detections
16.7
50.6
±12.5
56.517.6% 43.4% 22,213255,0301,407 (25.7)2,079 (37.9)2.6Public
Anonymous submission
EDMT17
16. using public detections
15.0
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
17. using public detections
13.8
49.9
±12.2
50.322.1% 36.7% 37,060243,1482,426 (42.6)3,846 (67.6)59.2Public
Anonymous submission
CEMT
18. using public detections
17.0
49.3
±12.6
44.416.8% 38.5% 21,711261,8082,696 (50.3)3,409 (63.6)5.8Public
Anonymous submission
TEM
19. using public detections
20.8
49.1
±12.6
45.417.0% 38.3% 22,119261,7973,439 (64.2)3,881 (72.4)8.2Public
Anonymous submission
PHD_PM_OM
20. online method using public detections
23.5
48.8
±13.4
43.219.1% 35.2% 26,260257,9714,407 (81.2)6,448 (118.8)0.6Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
ts_WCFMT
21. online method using public detections
20.8
48.4
±13.6
51.421.0% 32.5% 32,037255,4723,410 (62.3)6,351 (116.1)1.0Public
Anonymous submission
PHD_GSDL17
22. online method using public detections
20.3
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
23. online method using public detections
20.1
47.7
±13.0
48.916.5% 37.7% 18,991272,8423,534 (68.4)12,635 (244.7)24.9Public
Anonymous submission
WCFMT17
24. using public detections
20.7
47.3
±16.0
52.321.9% 30.7% 43,253250,3023,556 (63.9)6,071 (109.1)1.0Public
Anonymous submission
CFWM
25. online method using public detections
21.0
47.0
±13.3
42.718.1% 36.3% 26,123268,1874,649 (88.6)7,015 (133.7)9.7Public
Anonymous submission
SNM17
26. online method using public detections
28.1
46.8
±13.8
43.416.2% 37.1% 25,104271,0424,213 (81.1)9,891 (190.3)0.8Public
Anonymous submission
SSOMOT
27. online method using public detections
23.0
46.8
±13.1
49.215.3% 39.1% 24,041274,2572,121 (41.3)4,897 (95.3)4.9Public
Anonymous submission
COMOT
28. online method using public detections
19.8
46.4
±13.5
48.514.8% 42.2% 20,752279,8162,069 (41.0)4,606 (91.4)5.0Public
Anonymous submission
RTac
29. online method using public detections new
21.2
46.3
±14.6
49.218.9% 33.5% 43,447255,1584,196 (76.6)6,056 (110.6)14.1Public
Anonymous submission
IDOHMPT
30. online method using public detections
25.4
46.0
±13.1
44.116.8% 36.6% 30,873268,2215,768 (109.9)9,663 (184.2)8.1Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
MTS_CNN
31. online method using public detections
26.7
45.5
±12.7
41.417.3% 36.9% 33,774269,4934,033 (77.2)6,643 (127.2)2.8Public
Anonymous submission
IOU17
32. using public detections
25.7
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.
YT_T
33. online method using public detections
28.2
45.4
±13.4
40.316.0% 35.7% 25,425275,0507,652 (149.3)8,249 (160.9)11.4Public
Anonymous submission
LM_NN
34. using public detections
23.4
45.1
±13.3
43.214.8% 46.2% 10,834296,4512,286 (48.2)2,463 (51.9)2.5Public
JIVP-D-18-00047R1
FPSN
35. online method using public detections
25.3
44.9
±13.9
48.416.5% 35.8% 33,757269,9527,136 (136.8)14,491 (277.8)10.1Public
Anonymous submission
MTS_CNN_ML
36. online method using public detections
29.8
44.8
±12.7
37.717.1% 37.6% 30,661276,8363,798 (74.6)7,085 (139.1)0.4Public
Anonymous submission
HCC
37. using public detections
19.5
44.8
±11.2
46.818.3% 38.9% 17,586292,2941,555 (32.3)2,221 (46.1)0.9Public
Anonymous submission
reID2track
38. online method using public detections
30.6
44.6
±14.3
39.915.8% 39.7% 22,451284,2136,134 (123.6)13,786 (277.8)9.0Public
Anonymous submission
ReDetPast
39. online method using public detections
31.2
44.3
±14.8
34.917.3% 36.7% 32,113271,34310,962 (211.2)11,733 (226.0)3.3Public
Anonymous submission
STDIC
40. online method using public detections new
26.3
44.1
±13.6
45.913.2% 39.6% 46,126266,4492,992 (56.7)5,143 (97.4)17,757.0Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
Q_ls
41. online method using public detections
32.5
44.0
±13.7
40.115.7% 39.0% 41,442269,1495,136 (98.2)6,519 (124.7)0.6Public
Anonymous submission
RTRC
42. online method using public detections new
28.8
43.9
±14.2
47.413.1% 45.8% 34,309279,0302,986 (59.1)5,402 (106.9)9.8Public
Anonymous submission
ERC
43. online method using public detections
24.8
43.8
±13.1
49.213.8% 44.5% 27,729286,9172,299 (46.8)5,671 (115.4)5.9Public
Anonymous submission
PeriodMOT
44. online method using public detections
28.0
43.8
±13.2
40.914.7% 42.0% 21,941290,1944,910 (101.1)6,649 (136.9)66.9Public
Anonymous submission
GMPHD_SHA
45. online method using public detections
28.3
43.7
±12.5
39.211.7% 43.0% 25,935287,7583,838 (78.3)5,056 (103.2)9.2Public
Anonymous submission
EAMTT
46. online method using public detections
30.2
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.
ATIOU
47. online method using public detections
27.2
42.4
±15.0
40.312.9% 44.9% 12,720307,5804,804 (105.6)12,521 (275.3)123.3Public
Anonymous submission
MPTV1
48. online method using public detections
34.4
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
49. online method using public detections
34.3
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
50. online method using public detections
31.7
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.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
terry_T
51. online method using public detections new
37.8
2.5
±6.9
12.10.4% 83.9% 96,372448,0055,756 (279.4)11,270 (547.1)34.7Public
Anonymous submission
BIG_HA
52. online method using public detections new
27.1
-37.9
±24.1
0.00.0% 100.0% 213,867564,2280 (nan)0 (nan)887.9Public
Anonymous submission
CDT
53. using public detections new
27.5
-64.5
±16.9
0.10.0% 99.4% 364,642563,67272 (730.7)64 (649.5)46.9Public
Anonymous submission

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
21177572355564228

Difficulty Analysis

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

MOT17-03-SDP

MOT17-03-SDP

(67.1% MOTA)

MOT17-03-FRCNN

MOT17-03-FRCNN

(53.6% MOTA)

MOT17-06-SDP

MOT17-06-SDP

(47.5% MOTA)

...

...

MOT17-14-DPM

MOT17-14-DPM

(17.3% MOTA)

MOT17-14-FRCNN

MOT17-14-FRCNN

(16.8% 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.