2D MOT 2015 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 RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
APRCNN_Pub
1. online method using public detections
14.4
38.5
±9.9
72.60.78.7% 37.4% 4,00533,203586 (12.8)1,263 (27.5)6.7Public
Anonymous submission
MDPNN
2. online method using public detections
19.0
36.6
±12.1
71.41.113.3% 36.5% 6,41931,811700 (14.5)1,458 (30.2)1.9Public
A. Sadeghian, A. Alahi, S. Savarese. Tracking The Untrackable: Learning To Track Multiple Cues with Long-Term Dependencies. In arXiv preprint arXiv:1701.01909, 2017.
JointMC
3. using public detections
19.8
35.6
±18.9
71.91.823.2% 39.3% 10,58028,508457 (8.5)969 (18.1)0.6Public
PRMOT_DL61
4. online method using public detections
20.0
35.6
±14.5
69.81.417.8% 40.5% 7,91531,303338 (6.9)1,066 (21.7)2.2Public
Anonymous submission
mLK
5. online method using public detections
16.0
35.1
±12.9
71.51.012.3% 38.3% 5,67833,815383 (8.5)1,175 (26.1)1.0Public
Yuan Zhang, Di Xie and Shiliang Pu (Hikvision Research Institute)
TSMLCDEnew
6. using public detections
16.8
34.3
±13.1
71.71.414.0% 39.4% 7,86931,908618 (12.9)959 (20.0)6.5Public
B. Wang, G. Wang, K. L. Chan, L. Wang. Tracklet Association by Online Target-Specific Metric Learning and Coherent Dynamics Estimation. In arXiv:1511.06654, 2015.
QuadMOT
7. using public detections
19.8
33.8
±14.8
73.41.412.9% 36.9% 7,89832,061703 (14.7)1,430 (29.9)3.7Public
Anonymous submission
NOMT
8. using public detections
15.3
33.7
±16.2
71.91.312.2% 44.0% 7,76232,547442 (9.4)823 (17.5)11.5Public
W. Choi. Near-Online Multi-target Tracking with Aggregated Local Flow Descriptor. In ICCV, 2015.
TDAM
9. online method using public detections
20.3
33.0
±9.8
72.81.713.3% 39.1% 10,06430,617464 (9.2)1,506 (30.0)5.9Public
Y. Min, J. Yunde. Temporal Dynamic Appearance Modeling for Online Multi-Person Tracking. In arXiv:1510.02906, 2015.
omdpsplit
10. using public detections
23.8
32.6
±14.3
71.51.716.0% 34.4% 10,10530,704580 (11.6)1,268 (25.3)0.1Public
Anonymous submission
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
HG
11. online method using public detections
30.5
32.4
±12.2
70.31.18.9% 34.5% 6,48934,176882 (19.9)2,496 (56.2)0.5Public
CVPR #469
MHT_DAM
12. using public detections
19.8
32.4
±15.6
71.81.616.0% 43.8% 9,06432,060435 (9.1)826 (17.3)0.7Public
C. Kim, F. Li, A. Ciptadi, J. Rehg. Multiple Hypothesis Tracking Revisited. In ICCV, 2015.
CCF_MOT
13. online method using public detections
20.8
31.6
±14.2
71.81.010.1% 46.3% 6,06935,444491 (11.6)994 (23.5)0.9Public
icip #2242 Online Multiple Object Tracking via Flow and Convolutional Features
CIDAAT
14. online method using public detections
33.6
31.5
±15.8
70.72.112.9% 30.2% 12,29028,4411,354 (25.2)2,932 (54.6)1.2Public
cvpr #1109.
CF_MCMC
15. using public detections
31.0
31.4
±11.3
69.81.510.3% 40.9% 8,79832,541814 (17.3)1,711 (36.4)3.2Public
Anonymous submission
OAB
16. online method using public detections
24.3
31.2
±14.8
70.81.09.8% 46.3% 5,91635,974391 (9.4)1,115 (26.9)0.8Public
Anonymous submission
MPTCNN
17. online method using public detections
28.5
30.9
±12.9
70.11.39.4% 43.7% 7,37434,623484 (11.1)1,582 (36.2)0.6Public
Anonymous submission
TSSRC_2015
18. online method using public detections
25.8
30.5
±17.1
71.01.814.6% 41.1% 10,65031,445612 (12.5)1,585 (32.5)34.3Public
Anonymous submission
MDP
19. online method using public detections
27.9
30.3
±14.6
71.31.713.0% 38.4% 9,71732,422680 (14.4)1,500 (31.8)1.1Public
Y. Xiang, A. Alahi, S. Savarese. Learning to Track: Online Multi-Object Tracking by Decision Making. In International Conference on Computer Vision (ICCV), 2015.
OMDP15
20. using public detections
30.3
30.3
±14.4
71.11.915.3% 34.1% 11,09331,057682 (13.8)1,635 (33.1)0.1Public
Anonymous submission
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
DSR
21. using public detections
19.3
29.8
±13.1
72.21.09.7% 54.5% 5,69237,153269 (6.8)688 (17.4)2.6Public
Anonymous submission
CNNTCM
22. using public detections
23.6
29.6
±13.9
71.81.311.2% 44.0% 7,78634,733712 (16.4)943 (21.7)1.7Public
B. Wang, K. L. Chan, L. Wang, B. Shuai, Z. Zuo, T. Liu, G. Wang. Joint Learning of Convolutional Neural Networks and Temporally Constrained Metrics for Tracklet Association. In DeepVision Workshop (CVPR), 2016.
SCEA
23. online method using public detections
24.7
29.1
±12.2
71.11.08.9% 47.3% 6,06036,912604 (15.1)1,182 (29.6)6.8Public
J. Yoon, C. Lee, M. Yang, K. Yoon. Online Multi-object Tracking via Structural Constraint Event Aggregation. In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2016.
SiameseCNN
24. using public detections
23.6
29.0
±15.1
71.20.98.5% 48.4% 5,16037,798639 (16.6)1,316 (34.2)52.8Public
Laura Leal-Taixé, Cristian Canton-Ferrer, Konrad Schindler. Learning by Tracking: Siamese CNN for Robust Target Association. DeepVision Workshop (CVPR), Las Vegas (Nevada, USA), June 2016.
TBX
25. using public detections
34.8
27.5
±13.3
70.61.410.4% 45.8% 7,96835,810759 (18.2)1,528 (36.6)0.1Public
R. Henschel, L. Leal-Taixé, B. Rosenhahn, K. Schindler. Tracking with multi-level features. In arXiv:1607.07304, 2016.
oICF
26. online method using public detections
32.7
27.1
±14.9
70.01.36.4% 48.7% 7,59436,757454 (11.3)1,660 (41.3)1.4Public
H. Kieritz, S. Becker, W. Hübner, M. Arens. Online Multi-Person Tracking using Integral Channel Features. In IEEE Advanced Video and Signal-based Surveillance (AVSS) 2016, 2016.
LP_SSVM
27. using public detections
27.0
25.2
±13.7
71.71.45.8% 53.0% 8,36936,932646 (16.2)849 (21.3)41.3Public
S. Wang, C. Fowlkes. Learning Optimal Parameters for Multi-target Tracking with Contextual Interactions. In International Journal of Computer Vision, 2016.
ELP
28. using public detections
32.0
25.0
±10.8
71.21.37.5% 43.8% 7,34537,3441,396 (35.6)1,804 (46.0)5.7Public
N. McLaughlin, J. Martinez Del Rincon, P. Miller. Enhancing Linear Programming with Motion Modeling for Multi-target Tracking. In IEEE Winter Conference on Applications of Computer Vision (WACV), 2015.
LINF1
29. using public detections
25.7
24.5
±15.4
71.31.05.5% 64.6% 5,86440,207298 (8.6)744 (21.5)7.5Public
L. Fagot-Bouquet, R. Audigier, Y. Dhome, F. Lerasle. Improving Multi-Frame Data Association with Sparse Representations for Robust Near-Online Multi-Object Tracking. In ECCV, 2016.
MPO
30. online method using public detections
35.7
24.3
±11.1
70.90.83.1% 48.7% 4,72340,3401,459 (42.5)2,209 (64.3)11.3Public
Anonymous submission
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
JPDA_m
31. using public detections
30.0
23.8
±15.1
68.21.15.0% 58.1% 6,37340,084365 (10.5)869 (25.0)32.6Public
H. Rezatofighi, A. Milan, Z. Zhang, Q. Shi, A. Dick, I. Reid. Joint Probabilistic Data Association Revisited. In ICCV, 2015.
INTER_MOT
32. using public detections
33.1
23.6
±15.5
69.51.66.9% 51.9% 9,51937,081362 (9.1)1,007 (25.4)6.0Public
Anonymous submission
MTTCNN
33. using public detections new
33.0
23.5
±15.3
71.40.92.6% 54.0% 5,44840,802739 (22.0)1,208 (36.0)1.9Public
Anonymous submission
SFT_ACF
34. online method using public detections
38.2
23.2
±14.1
71.11.65.5% 43.0% 9,29536,5271,363 (33.6)2,363 (58.3)2.3Public
Anonymous submission
MotiCon
35. using public detections
39.1
23.1
±16.4
70.91.84.7% 52.0% 10,40435,8441,018 (24.4)1,061 (25.5)1.4Public
L. Leal-Taixé, M. Fenzi, A. Kuznetsova, B. Rosenhahn, S. Savarese. Learning an image-based motion context for multiple people tracking. In CVPR, 2014.
OLTT
36. using public detections
32.1
23.0
±12.6
71.21.33.3% 52.3% 7,39439,254644 (17.8)1,232 (34.1)10.9Public
Anonymous submission
Otakudj
37. online method using public detections
34.2
23.0
±14.0
70.72.217.8% 19.6% 12,76729,9144,625 (90.1)3,524 (68.7)11.6Public
Anonymous submission
SegTrack
38. using public detections
34.6
22.5
±15.2
71.71.45.8% 63.9% 7,89039,020697 (19.1)737 (20.2)0.2Public
A. Milan, L. Leal-Taixé, K. Schindler, I. Reid. Joint Tracking and Segmentation of Multiple Targets. In CVPR, 2015.
EAMTTpub
39. online method using public detections
36.9
22.3
±14.2
70.81.45.4% 52.7% 7,92438,982833 (22.8)1,485 (40.6)12.2Public
R. Sanchez-Matilla, F. Poiesi, A. Cavallaro "Multi-target tracking with strong and weak detections" in BMTT ECCVw 2016
pb
40. using public detections
36.6
21.3
±50.7
70.51.67.8% 54.6% 9,49637,999834 (21.9)1,423 (37.3)680.4Public
Anonymous submission
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
TFMOT
41. online method using public detections
26.5
21.3
±11.9
71.60.63.5% 67.8% 3,41244,595349 (12.7)654 (23.9)11.3Public
Anonymous submission
OMT_DFH
42. online method using public detections
36.0
21.2
±17.2
69.92.37.1% 46.5% 13,21834,657563 (12.9)1,255 (28.8)28.6Public
J. Ju, D. Kim, B. Ku, D. Han, H. Ko. Online multi-object tracking with efficient track drift and fragmentation handling. In J. Opt. Soc. Am. A, 2017.
simRNN
43. using public detections
43.5
20.3
±12.0
71.01.65.0% 54.8% 9,40638,588966 (26.0)1,464 (39.4)0.4Public
Anonymous submission
LP2D
44. using public detections
36.4
19.8
±14.2
71.22.06.7% 41.2% 11,58036,0451,649 (39.9)1,712 (41.4)112.1Public
MOT baseline: Linear programming on 2D image coordinates.
DCO_X
45. using public detections
38.6
19.6
±14.1
71.41.85.1% 54.9% 10,65238,232521 (13.8)819 (21.7)0.3Public
A. Milan, K. Schindler, S. Roth. Multi-Target Tracking by Discrete-Continuous Energy Minimization. In IEEE PAMI, 2016.
CEM
46. using public detections
39.9
19.3
±17.5
70.72.58.5% 46.5% 14,18034,591813 (18.6)1,023 (23.4)1.1Public
A. Milan, S. Roth, K. Schindler. Continuous Energy Minimization for Multitarget Tracking. In IEEE TPAMI, 2014.
RNN_LSTM
47. online method using public detections
39.7
19.0
±15.2
71.02.05.5% 45.6% 11,57836,7061,490 (37.0)2,081 (51.7)165.2Public
A. Milan, S. Rezatofighi, A. Dick, I. Reid, K. Schindler. Online Multi-Target Tracking using Recurrent Neural Networks. In AAAI, 2017.
RMOT
48. online method using public detections
42.4
18.6
±17.5
69.62.25.3% 53.3% 12,47336,835684 (17.1)1,282 (32.0)7.9Public
J. Yoon, H. Yang, J. Lim, K. Yoon. Bayesian Multi-Object Tracking Using Motion Context from Multiple Objects. In IEEE Winter Conference on Applications of Computer Vision (WACV), 2015.
TSDA_OAL
49. online method using public detections
38.9
18.6
±17.6
69.72.89.4% 42.3% 16,35032,853806 (17.3)1,544 (33.2)19.7Public
H. Ko. Online multi-person tracking with two-stage data association and online appearance model learning. In IET Computer Vision, 2017.
GMPHD_15
50. online method using public detections
35.3
18.5
±12.7
70.91.43.9% 55.3% 7,86441,766459 (14.3)1,266 (39.5)19.8Public
Y. Song, M. Jeon. Online Multiple Object Tracking with the Hierarchically Adopted GM-PHD Filter using Motion and Appearance. In IEEE/IEIE The International Conference on Consumer Electronics (ICCE) Asia, 2016.
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
SMOT
51. using public detections
43.8
18.2
±10.3
71.21.52.8% 54.8% 8,78040,3101,148 (33.4)2,132 (62.0)2.7Public
C. Dicle, O. Camps, M. Sznaier. The Way They Move: Tracking Targets with Similar Appearance. In ICCV, 2013.
ALExTRAC
52. using public detections
43.5
17.0
±12.1
71.21.63.9% 52.4% 9,23339,9331,859 (53.1)1,872 (53.5)3.7Public
A. Bewley, L. Ott, F. Ramos, B. Upcroft. ALExTRAC: Affinity Learning by Exploring Temporal Reinforcement within Association Chains. In International Conference on Robotics and Automation (ICRA), (to appear) 2016.
GMCSS
53. online method using public detections
47.4
16.7
±14.7
69.22.24.0% 55.1% 12,54338,102562 (14.8)1,619 (42.6)1.5Public
Anonymous submission
TBD
54. using public detections
47.5
15.9
±17.6
70.92.66.4% 47.9% 14,94334,7771,939 (44.7)1,963 (45.2)0.7Public
A. Geiger, M. Lauer, C. Wojek, C. Stiller, R. Urtasun. 3D Traffic Scene Understanding from Movable Platforms. In Pattern Analysis and Machine Intelligence (PAMI), 2014.
GSCR
55. online method using public detections
38.0
15.8
±10.5
69.41.31.8% 61.0% 7,59743,633514 (17.7)1,010 (34.8)28.1Public
L. Fagot-Bouquet, R. Audigier, Y. Dhome, F. Lerasle. Online multi-person tracking based on global sparse collaborative representations. In ICIP, 2015.
TC_ODAL
56. online method using public detections
48.8
15.1
±15.0
70.52.23.2% 55.8% 12,97038,538637 (17.1)1,716 (46.0)1.7Public
S. Bae, K. Yoon. Robust Online Multi-Object Tracking based on Tracklet Confidence and Online Discriminative Appearance Learning. In CVPR, 2014.
DP_NMS
57. using public detections
41.9
14.5
±13.9
70.82.36.0% 40.8% 13,17134,8144,537 (104.7)3,090 (71.3)444.8Public
H. Pirsiavash, D. Ramanan, C. Fowlkes. Globally-Optimal Greedy Algorithms for Tracking a Variable Number of Objects. In CVPR, 2011.
LP2D_REID
58. using public detections
41.5
13.6
±14.9
71.91.62.6% 66.2% 9,15640,6723,231 (95.6)1,407 (41.6)18.1Public
Anonymous submission
MOBS
59. using public detections
43.6
13.4
±16.2
71.51.82.6% 59.4% 10,29641,7801,123 (35.1)1,859 (58.1)2,891.5Public
Anonymous submission
FOT
60. online method using public detections
32.6
10.8
±17.0
71.42.115.4% 29.7% 12,05631,35711,373 (232.3)3,341 (68.2)96.4Public
Anonymous submission
TrackerAvg RankMOTAMOTPFAFMTMLFPFNID Sw.FragHzDetector
LDCT
61. online method using public detections
36.0
4.7
±41.3
71.72.411.4% 32.5% 14,06632,15612,348 (259.1)2,918 (61.2)20.7Public
F. Solera, S. Calderara, R. Cucchiara. Learning to Divide and Conquer for Online Multi-Target Tracking. In ICCV, 2015
cuMOT
62. online method using public detections new
26.6
0.8
±13.6
72.60.00.3% 98.3% 16160,72136 (30.8)62 (53.0)28.9Public
Anonymous submission

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
11578372161440

Difficulty Analysis

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

TUD-Crossing

TUD-Crossing

(61.9% MOTA)

PETS09-S2L2

PETS09-S2L2

(41.5% MOTA)

ETH-Jelmoli

ETH-Jelmoli

(36.4% MOTA)

...

...

KITTI-19

KITTI-19

(21.3% MOTA)

ADL-Rundle-1

ADL-Rundle-1

(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.
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 [2].
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] 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.