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 Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
CDT
1. using public detections
41.3
-64.5
±16.9
0.10.0% 99.4% 364,642563,67272 (730.7)64 (649.5)46.9Public
Anonymous submission
BIG_HA
2. online method using public detections
40.4
-37.9
±24.1
0.00.0% 100.0% 213,867564,2280 (nan)0 (nan)887.9Public
Anonymous submission
terry_T
3. online method using public detections
57.8
2.5
±6.9
12.10.4% 83.9% 96,372448,0055,756 (279.4)11,270 (547.1)34.7Public
Anonymous submission
GM_PHD
4. online method using public detections
49.2
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
5. online method using public detections
53.2
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.
EAMTT
6. online method using public detections
48.8
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.
BU_CV
7. online method using public detections
45.3
42.8
±14.4
32.315.8% 36.1% 40,573271,83810,118 (195.2)9,426 (181.9)17.8Public
Anonymous submission
ZM
8. online method using public detections
52.8
43.5
±13.9
32.614.5% 39.9% 25,083284,4059,197 (185.4)8,849 (178.4)14.4Public
Anonymous submission
PeriodMOT
9. online method using public detections
45.8
43.8
±13.2
40.914.7% 42.0% 21,941290,1944,910 (101.1)6,649 (136.9)66.9Public
Anonymous submission
REQT
10. online method using public detections
44.3
43.9
±14.2
47.413.1% 45.8% 34,309279,0302,986 (59.1)5,402 (106.9)64.1Public
Anonymous submission
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
STDIC
11. online method using public detections
42.5
44.1
±13.6
45.913.2% 39.6% 46,126266,4492,992 (56.7)5,143 (97.4)17,757.0Public
Anonymous submission
SAS_MOT17
12. using public detections
36.0
44.2
±12.2
57.216.1% 44.3% 29,473283,6111,529 (30.7)2,644 (53.2)4.8Public
Anonymous submission
ReDetPast
13. online method using public detections
49.4
44.3
±14.8
34.917.3% 36.7% 32,113271,34310,962 (211.2)11,733 (226.0)3.3Public
Anonymous submission
TM_track
14. online method using public detections
51.6
44.4
±14.6
43.915.8% 38.5% 26,067279,8328,038 (159.5)14,105 (279.8)2.5Public
Anonymous submission
reID2track
15. online method using public detections
48.3
44.6
±14.3
39.915.8% 39.7% 22,451284,2136,134 (123.6)13,786 (277.8)9.0Public
Anonymous submission
HCC
16. using public detections
31.9
44.8
±11.2
46.818.3% 38.9% 17,586292,2941,555 (32.3)2,221 (46.1)0.9Public
Anonymous submission
L_SORT
17. using public detections
40.8
45.0
±14.0
46.012.2% 41.1% 19,967287,2293,294 (67.1)8,292 (168.9)102.6Public
Anonymous submission
DSA_MOT17
18. online method using public detections
35.0
45.0
±12.6
43.615.8% 39.2% 21,442286,4822,491 (50.6)3,824 (77.7)9.9Public
Anonymous submission
dcor
19. online method using public detections
43.2
45.0
±14.2
34.015.4% 38.2% 30,231275,2654,801 (93.7)8,498 (165.9)44.4Public
Anonymous submission
LM_NN
20. using public detections
37.4
45.1
±13.3
43.214.8% 46.2% 10,834296,4512,286 (48.2)2,463 (51.9)2.5Public
NEUCOM-D-18-03230
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
YT_T
21. online method using public detections
46.3
45.4
±13.4
40.316.0% 35.7% 25,425275,0507,652 (149.3)8,249 (160.9)11.4Public
Anonymous submission
IOU17
22. using public detections
41.6
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.
EDA_GNN
23. online method using public detections
39.9
45.5
±13.8
40.515.6% 40.6% 25,685277,6634,091 (80.5)5,579 (109.8)39.3Public
Paper ID 2713
IDOHMPT
24. online method using public detections
42.1
46.0
±13.1
44.116.8% 36.6% 30,873268,2215,768 (109.9)9,663 (184.2)8.1Public
Anonymous submission
RTac
25. online method using public detections
35.8
46.3
±14.6
49.218.9% 33.5% 43,447255,1584,196 (76.6)6,056 (110.6)14.1Public
Anonymous submission
COMOT
26. online method using public detections
33.6
46.4
±13.5
48.514.8% 42.2% 20,752279,8162,069 (41.0)4,606 (91.4)5.0Public
Anonymous submission
SSOMOT
27. online method using public detections
37.8
46.8
±13.1
49.215.3% 39.1% 24,041274,2572,121 (41.3)4,897 (95.3)4.9Public
Anonymous submission
SNM17
28. online method using public detections
46.5
46.8
±13.8
43.416.2% 37.1% 25,104271,0424,213 (81.1)9,891 (190.3)0.8Public
Anonymous submission
QiMOT
29. online method using public detections
43.2
47.2
±13.1
40.815.5% 39.9% 18,907274,8284,320 (84.2)5,917 (115.4)1.8Public
Anonymous submission
WCFMT17
30. using public detections
35.6
47.3
±16.0
52.321.9% 30.7% 43,253250,3023,556 (63.9)6,071 (109.1)1.0Public
Anonymous submission
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
MHT_bLSTM
31. using public detections
32.6
47.5
±12.6
51.918.2% 41.7% 25,981268,0422,069 (39.4)3,124 (59.5)1.9Public
C. Kim, F. Li, J. Rehg. Multi-object Tracking with Neural Gating Using Bilinear LSTM. In ECCV, 2018.
PHD_GSDL17
32. online method using public detections
36.6
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.
AEb
33. using public detections
27.0
48.1
±13.4
46.017.7% 39.5% 16,839273,8192,350 (45.7)5,275 (102.5)66.9Public
Anonymous submission
AM_ADM17
34. online method using public detections
31.0
48.1
±13.8
52.113.4% 39.7% 25,061265,4952,214 (41.8)5,027 (94.9)5.7Public
S. Lee, M. Kim, S. Bae, Learning Discriminative Appearance Models for Online Multi-Object Tracking with Appearance Discriminability Measures, In IEEE Access, 2018.
DMAN
35. online method using public detections
30.2
48.2
±12.3
55.719.3% 38.3% 26,218263,6082,194 (41.2)5,378 (100.9)0.3Public
J. Zhu, H. Yang, N. Liu, M. Kim, W. Zhang, M. Yang. Online Multi-Object Tracking with Dual Matching Attention Networks. In ECCV, 2018.
TCF
36. online method using public detections
37.0
48.3
±13.6
48.718.9% 35.1% 36,274252,0923,530 (63.8)6,390 (115.5)6.4Public
Anonymous submission
HAM_SADF17
37. online method using public detections
28.8
48.3
±13.2
51.117.1% 41.7% 20,967269,0381,871 (35.8)3,020 (57.7)5.0Public
Y. Yoon, A. Boragule, Y. Song, K. Yoon, M. Jeon. Online Multi-Object Tracking with Historical Appearance Matching and Scene Adaptive Detection Filtering. In IEEE AVSS, 2018.
ts_WCFMT
38. online method using public detections
35.9
48.4
±13.6
51.421.0% 32.5% 32,037255,4723,410 (62.3)6,351 (116.1)1.0Public
Anonymous submission
RTRC
39. online method using public detections
35.9
48.5
±14.2
48.618.7% 35.7% 34,180252,8593,490 (63.2)6,304 (114.2)9.8Public
Anonymous submission
PHD_PM_OM
40. online method using public detections
39.8
48.8
±13.4
43.219.1% 35.2% 26,260257,9714,407 (81.2)6,448 (118.8)0.6Public
Anonymous submission
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
TEM
41. using public detections
35.5
49.1
±12.6
45.417.0% 38.3% 22,119261,7973,439 (64.2)3,881 (72.4)8.2Public
Anonymous submission
NV_MC
42. using public detections
32.1
49.1
±13.9
45.719.0% 38.0% 16,850267,9232,446 (46.6)3,196 (60.9)0.3Public
Anonymous submission
CEMT
43. using public detections
30.3
49.3
±12.6
44.416.8% 38.5% 21,711261,8082,696 (50.3)3,409 (63.6)5.8Public
Anonymous submission
MTDF17
44. online method using public detections
37.8
49.6
±13.9
45.218.9% 33.1% 37,124241,7685,567 (97.4)9,260 (162.0)1.2Public
Anonymous submission
IDGA
45. using public detections
24.6
49.9
±12.2
50.322.1% 36.7% 37,060243,1482,426 (42.6)3,846 (67.6)59.2Public
Anonymous submission
EDMT17
46. using public detections
26.9
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.
Q_ls
47. online method using public detections
36.8
50.2
±14.4
43.619.7% 37.3% 23,143253,1514,414 (80.1)6,112 (110.9)1.8Public
Anonymous submission
MOT_BJ
48. online method using public detections
34.3
50.4
±12.3
51.018.2% 34.1% 30,911245,8313,296 (58.4)6,279 (111.3)0.0Public
Anonymous submission
TLMHT
49. using public detections
28.2
50.6
±12.5
56.517.6% 43.4% 22,213255,0301,407 (25.7)2,079 (37.9)2.6Public
H. Sheng, J. Chen, Y. Zhang, W. Ke, Z. Xiong, J. Yu. Iterative Multiple Hypothesis Tracking with Tracklet-level Association. In IEEE Transactions on Circuits and Systems for Video Technology, 2018.
TBNMF17
50. online method using public detections
28.3
50.6
±12.6
49.318.9% 39.2% 17,522258,9902,014 (37.2)4,432 (81.9)6.9Public
Anonymous submission
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
ORCtracker
51. online method using public detections
33.7
50.7
±13.7
43.117.0% 35.2% 20,440249,7918,069 (144.8)11,188 (200.8)3,760.7Public
C. Deniz Cicek(Cortexica Vision System)
MHT_DAM
52. using public detections
28.1
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.
MOTDT17
53. online method using public detections
25.6
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.
SRPN17
54. online method using public detections
29.1
51.0
±11.7
53.516.8% 35.1% 21,011252,8082,596 (47.0)5,981 (108.4)4.1Public
Anonymous submission
jCC
55. using public detections
22.8
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, B. Andres, T. Brox, B. Schiele. Motion Segmentation amp; Multiple Object Tracking by Correlation Co-Clustering. In IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018.
hpmmt17
56. online method using public detections
21.2
51.2
±11.8
53.617.3% 34.9% 21,957250,8912,292 (41.3)6,108 (110.0)44,392.5Public
Anonymous submission
Sn_PBC
57. using public detections
26.5
51.3
±11.7
53.417.4% 35.2% 21,255251,2562,394 (43.2)6,148 (110.8)14.8Public
Anonymous submission
FWT
58. using public detections
27.1
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.
TBD17_1
59. online method using public detections
26.8
51.4
±11.7
52.018.5% 33.2% 24,261247,1952,985 (53.1)6,611 (117.7)1,183.8Public
Anonymous submission
AFN17
60. using public detections
23.8
51.5
±13.0
46.920.6% 35.5% 22,391248,4202,593 (46.3)4,308 (77.0)1.8Public
Paper ID 4411
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
PA_MOT17
61. online method using public detections
22.4
51.6
±13.5
53.518.9% 33.5% 28,794241,7042,635 (46.1)5,808 (101.6)710.3Public
Anonymous submission
TAR_1
62. online method using public detections
30.7
51.6
±11.9
41.421.7% 28.7% 33,514235,8593,629 (62.4)5,949 (102.2)5.6Public
Anonymous submission
MOT_test
63. online method using public detections
23.5
51.6
±11.9
53.917.3% 35.5% 21,419249,0592,384 (42.7)5,613 (100.5)7.8Public
Anonymous submission
eHAF17
64. using public detections
20.9
51.8
±13.2
54.723.4% 37.9% 33,212236,7721,834 (31.6)2,739 (47.2)0.7Public
H. Sheng, Y. Zhang, J. Chen, Z. Xiong, J. Zhang. Heterogeneous Association Graph Fusion for Target Association in Multiple Object Tracking. In IEEE Transactions on Circuits and Systems for Video Technology, 2018.
TOPA
65. online method using public detections new
23.8
51.8
±13.5
53.419.6% 33.1% 27,603241,5462,668 (46.7)5,790 (101.2)443.9Public
Anonymous submission
eTC17
66. using public detections
20.7
51.9
±12.8
58.023.5% 35.5% 37,311231,6582,294 (38.9)2,917 (49.5)0.7Public
Anonymous submission
CMT
67. using public detections
15.7
52.0
±13.2
54.823.3% 37.7% 31,660237,5471,827 (31.6)2,738 (47.3)10.2Public
Anonymous submission
SemiOMOT
68. using public detections
23.3
52.4
±15.0
51.022.6% 34.6% 23,660242,9532,070 (36.4)3,170 (55.7)0.7Public
Anonymous submission
BnW
69. online method using public detections
23.0
52.5
±15.3
52.618.7% 36.7% 19,192245,7652,822 (50.0)5,610 (99.4)1.7Public
Anonymous submission
yt_face
70. online method using public detections
22.8
52.6
±13.1
51.523.0% 35.9% 23,894241,4892,047 (35.8)2,827 (49.4)2.2Public
Anonymous submission
TrackerAvg Rank MOTAIDF1MTMLFPFNID Sw.FragHzDetector
IOUT_Re
71. online method using public detections
26.5
52.7
±13.0
43.320.1% 32.6% 16,529243,2266,946 (122.1)6,520 (114.6)7.0Public
Anonymous submission
NOTBD
72. using public detections
26.2
53.9
±12.7
51.221.5% 35.6% 28,912228,3562,964 (49.8)3,600 (60.5)0.3Public
Anonymous submission
HIK_MOT17
73. using public detections
16.4
53.9
±13.7
54.323.7% 32.0% 27,656230,0422,386 (40.3)4,192 (70.8)5.4Public
Qclc
74. online method using public detections
25.5
54.0
±14.3
47.723.3% 30.7% 22,374232,2124,748 (80.7)6,022 (102.3)1.8Public
Anonymous submission
DH_TRK
75. using public detections
22.8
54.1
±13.0
49.221.6% 28.4% 36,196216,6705,918 (96.1)7,760 (126.0)1,775.7Public
Anonymous submission
HDTR
76. using public detections
16.0
54.1
±11.4
48.423.3% 34.8% 18,002238,8181,895 (32.9)2,693 (46.7)1.8Public
TPM
77. using public detections
20.6
54.2
±13.0
52.622.8% 37.5% 13,739242,7301,824 (32.0)2,472 (43.4)0.8Public
Anonymous submission
DGCT
78. using public detections
20.4
54.3
±13.1
50.620.3% 33.7% 15,508238,2464,243 (73.4)6,684 (115.7)7.0Public
CJY, HYW, KHW @ HRI-SH
ISDH_HDAv2
79. online method using public detections
21.3
54.5
±14.5
65.926.4% 32.1% 46,693207,0933,010 (47.6)6,000 (94.8)3.6Public
MM-008988/ IEEE Transactions on Multimedia

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
21177572355564228

Difficulty Analysis

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

MOT17-03-SDP

MOT17-03-SDP

(69.5% MOTA)

MOT17-03-FRCNN

MOT17-03-FRCNN

(56.3% MOTA)

MOT17-06-SDP

MOT17-06-SDP

(49.8% MOTA)

...

...

MOT17-14-DPM

MOT17-14-DPM

(20.6% MOTA)

MOT17-14-FRCNN

MOT17-14-FRCNN

(20.1% 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.