MOT20 Results

Click on a measure to sort the table accordingly. See below for a more detailed description.


Showing only entries that use public detections!


Benchmark Statistics

TrackerMOTAIDF1MOTPMTMLFPFNRecallPrecisionFAFID Sw.FragHz
SORT20
1. online method using public detections
42.7
±18.6
45.1
±13.1
78.5 208 (16.7)326 (26.2)27,521 264,694 48.8 90.2 6.1 4,470 (91.5)17,798 (364.4)57.3
A. Bewley, Z. Ge, L. Ott, F. Ramos, B. Upcroft. Simple online and realtime tracking. In 2016 IEEE International Conference on Image Processing (ICIP), 2016.
Surveily
2. online method using public detections
44.6
±24.3
42.5
±12.2
76.1 393 (31.6)236 (19.0)71,208 211,064 59.2 81.1 15.9 4,334 (73.2)6,646 (112.2)29.8
HOMI_Tracker
3. online method using public detections
51.2
±22.6
43.0
±13.3
79.6 423 (34.1)312 (25.1)16,094 232,259 55.1 94.7 3.6 3,937 (71.4)6,458 (117.2)7.5
BBT
4. using public detections
46.8
±21.3
42.2
±12.0
78.0 312 (25.1)289 (23.3)35,014 236,176 54.4 88.9 7.8 3,880 (71.4)7,207 (132.6)8.0
MTAP-D-20-01870 Tracking Subjects and Detecting Relationships in Crowded City Videos (under review)
SFS
5. online method using public detections
50.8
±17.4
41.1
±9.6
74.9 341 (27.5)251 (20.2)50,139 200,932 61.2 86.3 11.2 3,503 (57.3)7,617 (124.5)0.1
MTSFS´╝Ü Online Multi-Object Tracking Based on Salient Feature Selection in Crowded Scenes
UNS20
6. online method using public detections
47.8
±21.5
44.9
±9.8
78.1 335 (27.0)269 (21.7)39,587 227,328 56.1 88.0 8.8 3,067 (54.7)4,388 (78.3)24.9
Anonymous submission
MOT20_TBC
7. using public detections
54.5
±17.2
50.1
±11.5
77.3 415 (33.4)245 (19.7)37,937 195,242 62.3 89.5 8.5 2,449 (39.3)2,580 (41.4)5.6
BD_MOT
8. using public detections
53.5
±19.8
49.3
±13.5
79.8 383 (30.8)321 (25.8)7,211 230,862 55.4 97.5 1.6 2,336 (42.2)4,673 (84.4)9.4
Anonymous submission
UnsupTrack
9. online method using public detections
53.6
±19.0
50.6
±13.3
80.1 376 (30.3)311 (25.0)6,439 231,298 55.3 97.8 1.4 2,178 (39.4)4,335 (78.4)1.3
S. Karthik, A. Prabhu, V. Gandhi. Simple Unsupervised Multi-Object Tracking. In Arxiv, 2020.
LPC_MOT
10. using public detections
56.3
±18.9
62.5
±14.9
79.7 424 (34.1)313 (25.2)11,726 213,056 58.8 96.3 2.6 1,562 (26.6)1,865 (31.7)0.7
Tsinghua University & AIBEE Research.
SequencesFramesTrajectoriesBoxes
444791501765465

Difficulty Analysis

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

MOT20-04

MOT20-04

(69.4 MOTA)

MOT20-07

MOT20-07

(51.9 MOTA)

MOT20-06

MOT20-06

(28.5 MOTA)

MOT20-08

MOT20-08

(18.5 MOTA)


Evaluation Measures

Lower is better. Higher is better.
Measure Better Perfect Description
MOTA higher 100%Multi-Object Tracking Accuracy (+/- denotes standard deviation across all sequences) [1]. This measure combines three error sources: false positives, missed targets and identity switches.
IDF1 higher 100%ID F1 Score [2]. The ratio of correctly identified detections over the average number of ground-truth and computed detections.
MOTP higher 100%Multi-Object Tracking Precision (+/- denotes standard deviation across all sequences) [1]. The misalignment between the annotated and the predicted bounding boxes.
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 0The total number of false positives.
FN lower 0The total number of false negatives (missed targets).
Recall higher 100%Ratio of correct detections to total number of GT boxes.
Precision higher 100%Ratio of TP / (TP+FP).
FAF lower 0The average number of false alarms per frame.
ID Sw. lower 0Number of Identity Switches (ID switch ratio = #ID switches / recall) [3]. Please note that we follow the stricter definition of identity switches as described in the reference
Frag lower 0The 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. The frequency is provided by the authors and not officially evaluated by the MOTChallenge.

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.
using private detections This method used a private 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.