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
DH_TRK
1. using public detections
15.6
54.1
±13.0
49.221.6% 28.4% 36,196216,6705,918 (96.1)7,760 (126.0)1,775.7Public
Anonymous submission
HIK_MOT17
2. using public detections
11.0
53.9
±13.7
54.323.7% 32.0% 27,656230,0422,386 (40.3)4,192 (70.8)5.4Public
NOTBD
3. using public detections
17.8
53.9
±12.7
51.221.5% 35.6% 28,912228,3562,964 (49.8)3,600 (60.5)0.3Public
Anonymous submission
IOUT_Re
4. online method using public detections
18.3
52.7
±13.0
43.320.1% 32.6% 16,529243,2266,946 (122.1)6,520 (114.6)7.0Public
Anonymous submission
yt_face
5. online method using public detections
14.9
52.6
±13.1
51.523.0% 35.9% 23,894241,4892,047 (35.8)2,827 (49.4)2.2Public
Anonymous submission
SST
6. online method using public detections
21.7
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)
SemiOMOT
7. using public detections
15.2
52.4
±15.0
51.022.6% 34.6% 23,660242,9532,070 (36.4)3,170 (55.7)0.7Public
Anonymous submission
CNN_search
8. using public detections
13.6
52.2
±13.8
50.721.4% 33.3% 23,565243,2323,059 (53.8)4,702 (82.6)3.7Public
Anonymous submission
eHAF17
9. using public detections
13.8
51.8
±13.2
54.723.4% 37.9% 33,212236,7721,834 (31.6)2,739 (47.2)0.7Public
TCSVT-02141-2018
MOT_test
10. online method using public detections new
15.6
51.6
±11.9
53.917.3% 35.5% 21,419249,0592,384 (42.7)5,613 (100.5)7.8Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
PA_MOT17
11. online method using public detections
15.4
51.6
±13.5
53.518.9% 33.5% 28,794241,7042,635 (46.1)5,808 (101.6)710.3Public
Anonymous submission
AFN17
12. using public detections
15.9
51.5
±13.0
46.920.6% 35.5% 22,391248,4202,593 (46.3)4,308 (77.0)1.8Public
Paper ID 4411
FWT
13. using public detections
18.0
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
14. using public detections
15.3
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.
hpmmt17
15. online method using public detections
15.1
51.0
±11.7
53.416.8% 35.5% 20,695253,1412,548 (46.2)5,919 (107.4)44,392.5Public
Anonymous submission
MOTDT17
16. online method using public detections
17.3
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
17. using public detections
18.8
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.
ORCtracker
18. online method using public detections
23.0
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)
TLMHT
19. using public detections
19.3
50.6
±12.5
56.517.6% 43.4% 22,213255,0301,407 (25.7)2,079 (37.9)2.6Public
TCSVT-02160-2018
EDMT17
20. using public detections
17.4
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.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
IDGA
21. using public detections
16.3
49.9
±12.2
50.322.1% 36.7% 37,060243,1482,426 (42.6)3,846 (67.6)59.2Public
Anonymous submission
CEMT
22. using public detections
20.4
49.3
±12.6
44.416.8% 38.5% 21,711261,8082,696 (50.3)3,409 (63.6)5.8Public
Anonymous submission
NV_MC
23. using public detections
22.3
49.1
±13.9
45.719.0% 38.0% 16,850267,9232,446 (46.6)3,196 (60.9)0.3Public
Anonymous submission
PHD_PM_OM
24. online method using public detections
27.0
48.8
±13.4
43.219.1% 35.2% 26,260257,9714,407 (81.2)6,448 (118.8)0.6Public
Anonymous submission
RTRC
25. online method using public detections
24.3
48.5
±14.2
48.618.7% 35.7% 34,180252,8593,490 (63.2)6,304 (114.2)9.8Public
Anonymous submission
ts_WCFMT
26. online method using public detections
23.8
48.4
±13.6
51.421.0% 32.5% 32,037255,4723,410 (62.3)6,351 (116.1)1.0Public
Anonymous submission
HAM_SADF17
27. online method using public detections
19.3
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.
TCF
28. online method using public detections
25.1
48.3
±13.6
48.718.9% 35.1% 36,274252,0923,530 (63.8)6,390 (115.5)6.4Public
Anonymous submission
DMAN
29. online method using public detections
20.8
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.
PHD_GSDL17
30. online method using public detections
24.7
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.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
FLT
31. online method using public detections
24.1
47.7
±13.0
48.916.5% 37.7% 18,991272,8423,534 (68.4)12,635 (244.7)24.9Public
Anonymous submission
MHT_bLSTM
32. using public detections
21.8
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.
WCFMT17
33. using public detections
24.0
47.3
±16.0
52.321.9% 30.7% 43,253250,3023,556 (63.9)6,071 (109.1)1.0Public
Anonymous submission
SNM17
34. online method using public detections
32.7
46.8
±13.8
43.416.2% 37.1% 25,104271,0424,213 (81.1)9,891 (190.3)0.8Public
Anonymous submission
SSOMOT
35. online method using public detections
26.8
46.8
±13.1
49.215.3% 39.1% 24,041274,2572,121 (41.3)4,897 (95.3)4.9Public
Anonymous submission
COMOT
36. online method using public detections
23.1
46.4
±13.5
48.514.8% 42.2% 20,752279,8162,069 (41.0)4,606 (91.4)5.0Public
Anonymous submission
RTac
37. online method using public detections
25.0
46.3
±14.6
49.218.9% 33.5% 43,447255,1584,196 (76.6)6,056 (110.6)14.1Public
Anonymous submission
IDOHMPT
38. online method using public detections
29.3
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
39. online method using public detections
30.8
45.5
±12.7
41.417.3% 36.9% 33,774269,4934,033 (77.2)6,643 (127.2)2.8Public
Anonymous submission
IOU17
40. using public detections
29.2
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.
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
YT_T
41. online method using public detections
32.8
45.4
±13.4
40.316.0% 35.7% 25,425275,0507,652 (149.3)8,249 (160.9)11.4Public
Anonymous submission
MTS_CNN_ML
42. online method using public detections
33.3
44.8
±12.7
37.717.1% 37.6% 30,661276,8363,798 (74.6)7,085 (139.1)0.4Public
Anonymous submission
HCC
43. using public detections
22.3
44.8
±11.2
46.818.3% 38.9% 17,586292,2941,555 (32.3)2,221 (46.1)0.9Public
Anonymous submission
TOPA
44. online method using public detections
28.8
44.8
±14.3
47.614.0% 43.5% 15,673293,7532,153 (44.9)7,142 (149.0)443.9Public
Anonymous submission
reID2track
45. online method using public detections
34.2
44.6
±14.3
39.915.8% 39.7% 22,451284,2136,134 (123.6)13,786 (277.8)9.0Public
Anonymous submission
ReDetPast
46. online method using public detections
35.1
44.3
±14.8
34.917.3% 36.7% 32,113271,34310,962 (211.2)11,733 (226.0)3.3Public
Anonymous submission
STDIC
47. online method using public detections
30.0
44.1
±13.6
45.913.2% 39.6% 46,126266,4492,992 (56.7)5,143 (97.4)17,757.0Public
Anonymous submission
Q_ls
48. online method using public detections
36.5
44.0
±13.7
40.115.7% 39.0% 41,442269,1495,136 (98.2)6,519 (124.7)0.6Public
Anonymous submission
REQT
49. online method using public detections
31.7
43.9
±14.2
47.413.1% 45.8% 34,309279,0302,986 (59.1)5,402 (106.9)64.1Public
Anonymous submission
PeriodMOT
50. online method using public detections
31.9
43.8
±13.2
40.914.7% 42.0% 21,941290,1944,910 (101.1)6,649 (136.9)66.9Public
Anonymous submission
TrackerAvg RankMOTAIDF1MTMLFPFNID Sw.FragHzDetector
GMPHD_SHA
51. online method using public detections
31.3
43.7
±12.5
39.211.7% 43.0% 25,935287,7583,838 (78.3)5,056 (103.2)9.2Public
Anonymous submission
EAMTT
52. online method using public detections
33.8
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.
GMPHD_KCF
53. online method using public detections
37.7
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
54. online method using public detections
34.8
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.
terry_T
55. online method using public detections
41.2
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
56. online method using public detections
29.2
-37.9
±24.1
0.00.0% 100.0% 213,867564,2280 (nan)0 (nan)887.9Public
Anonymous submission
CDT
57. using public detections
29.8
-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

(68.4% MOTA)

MOT17-03-FRCNN

MOT17-03-FRCNN

(54.5% MOTA)

MOT17-06-SDP

MOT17-06-SDP

(48.0% MOTA)

...

...

MOT17-14-DPM

MOT17-14-DPM

(18.9% MOTA)

MOT17-14-FRCNN

MOT17-14-FRCNN

(18.2% 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.