CVPR 2019 Tracking Challenge 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 RankMOTAIDF1MT MLFPFNID Sw.FragHzDetector
SRK_ODESA
1. online method using public detections
7.6
54.8
±19.3
52.235.4% 19.2% 33,814215,5723,750 (61.0)5,493 (89.3)1.2Public
D. Borysenko, D. Mykheievskyi, V. Porokhonskyy. ODESA: Object Descriptor that is Smooth Appearance-wise for object tracking tasks. In (to be submitted to ECCV'20), .
HAM_HI
2. online method using public detections
10.1
43.0
±24.0
43.628.1% 21.8% 72,018243,0554,153 (73.4)4,801 (84.8)0.8Public
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.
PA_MOT19V2
3. online method using public detections
11.2
44.4
±19.9
40.724.0% 22.1% 55,387250,7374,995 (90.4)8,468 (153.3)44.8Public
MOT19ZH
4. online method using public detections
13.3
44.6
±17.3
35.519.2% 22.7% 35,905268,8275,682 (109.3)7,767 (149.4)0.6Public
Anonymous submission
SiamMOT
5. online method using public detections
18.2
37.6
±17.1
27.617.7% 23.0% 64,916275,0729,709 (190.8)11,831 (232.5)0.2Public
Anonymous submission
BD_19
6. using public detections
6.8
45.5
±17.9
46.423.2% 23.3% 55,142247,2502,671 (47.8)3,281 (58.7)7.9Public
Anonymous submission
DD_TAMA19
7. online method using public detections
6.8
47.6
±20.3
48.727.2% 23.6% 38,194252,9342,437 (44.4)3,887 (70.9)0.2Public
Y. Yoon, D. Kim, K. Yoon, Y. Song, M. Jeon. Online Multiple Pedestrian Tracking using Deep Temporal Appearance Matching Association. In arXiv:1907.00831, 2019.
TrackReID
8. using public detections
14.0
41.7
±22.2
37.524.8% 23.8% 69,252253,0314,121 (75.2)5,371 (98.0)0.4Public
Anonymous submission
V_IOU
9. using public detections
8.7
46.7
±19.6
46.022.9% 24.4% 33,776261,9642,589 (48.6)4,354 (81.8)18.2Public
E. Bochinski, T. Senst, T. Sikora. Extending IOU Based Multi-Object Tracking by Visual Information. In IEEE International Conference on Advanced Video and Signals-based Surveillance, 2018.
DeepMOT19
10. online method using public detections
10.8
44.3
±19.5
37.423.4% 24.4% 46,152263,0372,740 (51.7)3,710 (70.0)0.5Public
Anonymous submission
TrackerAvg RankMOTAIDF1MT MLFPFNID Sw.FragHzDetector
Aaron
11. using public detections
6.2
46.5
±18.6
46.622.5% 24.6% 40,676256,6712,315 (42.7)2,968 (54.8)14.9Public
Anonymous submission
CASAVP3
12. online method using public detections
12.4
42.1
±20.1
43.820.6% 24.9% 49,141271,1404,027 (78.1)6,475 (125.5)0.5Public
Anonymous submission
CASAVP2
13. online method using public detections
12.3
42.4
±19.6
42.218.5% 25.2% 41,983276,8073,685 (72.9)6,232 (123.2)0.7Public
Anonymous submission
SAMOT_o
14. online method using public detections
13.7
43.9
±18.0
44.317.0% 25.9% 30,972280,0633,190 (63.8)7,775 (155.5)0.3Public
Anonymous submission
TracktorCV
15. online method using public detections
8.5
51.3
±18.7
47.624.9% 26.0% 16,263253,6802,584 (47.2)4,824 (88.2)0.6Public
P. Bergmann, T. Meinhardt, L. Leal-Taixe. Tracking without bells and whistles. In CoRR, 2019.
IITB_trk
16. online method using public detections
10.1
45.5
±15.4
43.618.7% 26.3% 23,931278,0423,002 (59.6)5,478 (108.8)2.5Public
Anonymous submission
SOTD_MCL19
17. online method using public detections
18.5
33.5
±17.1
24.811.8% 26.7% 71,644294,6176,177 (130.3)8,326 (175.7)16.9Public
Anonymous submission
NAR
18. using public detections
11.3
44.7
±17.1
42.018.6% 27.1% 25,857281,5192,527 (50.8)2,839 (57.1)0.6Public
Anonymous submission
S2ST
19. online method using public detections
16.7
40.0
±16.7
37.514.3% 27.3% 38,249292,7325,006 (104.9)9,914 (207.7)7.8Public
Anonymous submission
GNA
20. using public detections
9.8
44.8
±17.5
41.918.4% 27.6% 21,392285,4072,451 (50.0)2,699 (55.0)1.2Public
Anonymous submission
TrackerAvg RankMOTAIDF1MT MLFPFNID Sw.FragHzDetector
T_MHT19
21. using public detections
10.8
44.9
±17.0
49.221.8% 27.7% 34,808271,6352,343 (45.5)2,688 (52.2)0.2Public
Anonymous submission
DSORT_jpda
22. online method using public detections
16.7
39.1
±16.3
34.811.1% 29.0% 28,114307,8374,834 (107.3)12,634 (280.6)1.0Public
Anonymous submission
MEG_DeeS_A
23. online method using public detections
12.3
42.3
±16.4
42.714.2% 29.7% 20,253300,2612,473 (53.3)6,065 (130.8)3.0Public
Anonymous submission
DSTlpx19
24. online method using public detections
14.9
39.1
±16.1
37.511.1% 29.9% 25,528311,8563,863 (87.2)11,004 (248.3)22.4Public
Anonymous submission
MOTDT_19
25. online method using public detections
14.4
40.8
±14.2
39.110.4% 31.0% 12,079316,0523,292 (75.6)8,707 (199.9)22.4Public
Anonymous submission
IOU_19
26. using public detections
17.9
35.8
±15.1
25.710.0% 31.0% 24,427319,69615,676 (365.3)17,864 (416.3)183.3Public
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.
MGT
27. online method using public detections
14.9
39.1
±15.3
37.510.8% 31.6% 19,145318,1183,547 (82.1)10,430 (241.5)22,395.0Public
Anonymous submission
GX615_19_1
28. online method using public detections
18.5
30.1
±13.5
24.07.0% 35.8% 24,911350,94015,528 (415.9)18,412 (493.2)29,860.0Public
Anonymous submission
GH
29. using public detections
15.5
33.9
±13.5
26.77.1% 38.8% 15,477350,7544,020 (107.6)4,748 (127.1)8.3Public
Anonymous submission

Benchmark Statistics

SequencesFramesTrajectoriesBoxes
444791492803370

Difficulty Analysis

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

CVPR19-04

CVPR19-04

(56.0% MOTA)

CVPR19-07

CVPR19-07

(50.0% MOTA)

CVPR19-06

CVPR19-06

(24.5% MOTA)

CVPR19-08

CVPR19-08

(14.1% 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.