MOT Challenge - Results

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

Benchmark Statistics

Tracker	AP	MODA	MODP	FAF	TP	FP	FN	Recall	Precision	F1	Hz
DeDeT 1.	0.89 ±0.06	70.7 ±18.7	81.5	16.7	317,832	75,004	25,692	92.5	80.9	86.3	16.9
DeDeT 1.
Bresee_Det 2.	0.87 ±0.09	35.6 ±45.8	73.3	42.6	313,100	190,776	30,424	91.1	62.1	73.9	11.2
Bresee_Det 2.
GNN_SDT 3.	0.81 ±0.09	79.3 ±15.1	80.0	7.1	304,236	31,677	39,288	88.6	90.6	89.6	1.2
GNN_SDT 3.	Y. Wang, X. Weng, K. Kitani. Joint Detection and Multi-Object Tracking with Graph Neural Networks. In arXiv, 2020.
ViPeD20 4.	0.80 ±0.07	46.0 ±40.5	79.0	31.1	297,101	139,111	46,277	86.5	68.1	76.2	11.2
ViPeD20 4.	L. Ciampi, N. Messina, F. Falchi, C. Gennaro, G. Amato. Virtual to Real Adaptation of Pedestrian Detectors. In Sensors, 2020.

Sequences	Frames	Trajectories	Boxes
4	4479	1501	765465

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

MOT20-07

(0.90 AP)

MOT20-04

(0.88 AP)

MOT20-08

(0.76 AP)

MOT20-06

(0.73 AP)

Lower is better. Higher is better.

Measure	Better	Perfect	Description
AP	higher	1	Average Precision taken over a set of reference recall values (0:0.1:1)
MODA	higher	100%	Multi-Object Detection Accuracy [1]. This measure combines false positives and missed targets.
MODP	higher	100%	Multi-Object Detection Precision [1]. The misalignment between the annotated and the predicted bounding boxes.
FAF	lower	0	The average number of false alarms per frame.
TP	higher	#GT	The total number of true positives.
FP	lower	0	The total number of false positives.
FN	lower	0	The 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).
F1	higher	100%	Harmonic mean of precision and recall.
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.

Symbol	Description
	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.
	This method used the provided detection set as input.
	This method used a private detection set as input.
	This entry has been submitted or updated less than a week ago.

[1]	Bernardin, K. & Stiefelhagen, R. Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. Image and Video Processing, 2008(1):1-10, 2008.