MOT Challenge

Video not available.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Rendering of new sequences is currently deactivated due to heavy load.

Benchmark:

MOT15 | MOT16 | MOT17 | CVPR 2019 Tracking Challenge |

Short name:

Tracktor++

Detector:

Public

Description:

The problem of tracking multiple objects in a video sequence poses several challenging tasks. For tracking-by- detection these include object re-identification, motion pre- diction and dealing with occlusions. We present a tracker that accomplishes tracking without specifically targeting any of these tasks, in particular, we perform no training or optimization on tracking data. To this end, we exploit the bounding box regression of an object detector to predict the position of an object in the next frame, thereby converting a detector into a Tracktor. We demonstrate the extensibility of our Tracktor and provide a new state-of-the-art on three multi-object tracking benchmarks by extending it with a straightforward re-identification and camera motion compensation. This benchmark submission presents the results of our extended Tracktor++ multi-object tracker.

Reference:

P. Bergmann, T. Meinhardt, L. Leal-Taixé. Tracking without bells and whistles. In ICCV, 2019.

Last submitted:

April 10, 2019 (5 years ago)

Published:

April 10, 2019 at 17:29:58 CET

Submissions:

Project page / code:

https://github.com/phil-bergmann/tracking_wo_bnw

Open source:

Yes

Hardware:

Titan X 12 GB

Runtime:

1.5 Hz

Benchmark performance:

Sequence	MOTA	IDF1	HOTA	MT	ML	FP	FN	Rcll	Prcn	AssA	DetA	AssRe	AssPr	DetRe	DetPr	LocA	FAF	ID Sw.	Frag
MOT17	53.5	52.3	42.1	459 (19.5)	861 (36.6)	12,201	248,047	56.0	96.3	41.7	42.9	45.4	75.6	45.0	77.3	80.9	0.7	2,072 (37.0)	4,611 (82.3)

Detailed performance:

Sequence	MOTA	IDF1	HOTA	MT	ML	FP	FN	Rcll	Prcn	AssA	DetA	AssRe	AssPr	DetRe	DetPr	LocA	FAF	ID Sw.	Frag
MOT17-01-DPM	35.9	37.1	32.1	5	12	131	3,962	38.6	95.0	35.9	29.0	39.1	73.5	30.1	74.1	78.8	0.3	39	73
MOT17-01-FRCNN	34.9	34.8	31.4	5	10	406	3,753	41.8	86.9	32.6	30.7	37.8	68.9	32.9	68.5	77.8	0.9	39	87
MOT17-01-SDP	37.5	36.8	32.9	6	10	283	3,706	42.5	90.7	35.2	31.2	38.5	72.0	33.1	70.4	78.1	0.6	42	95
MOT17-03-DPM	65.2	57.0	45.8	52	19	1,338	34,840	66.7	98.1	41.1	51.6	44.7	75.8	53.9	79.2	81.6	0.9	222	343
MOT17-03-FRCNN	66.4	59.7	47.4	55	20	1,014	33,961	67.6	98.6	43.6	52.0	47.2	76.0	54.3	79.2	81.4	0.7	189	327
MOT17-03-SDP	69.6	60.1	48.7	59	16	2,469	29,065	72.2	96.8	43.5	55.1	47.5	74.3	58.0	77.8	80.8	1.6	248	525
MOT17-06-DPM	52.7	55.7	44.1	41	89	184	5,310	54.9	97.2	44.8	43.6	50.0	76.8	45.5	80.5	83.0	0.2	80	168
MOT17-06-FRCNN	56.7	59.0	46.2	51	61	359	4,647	60.6	95.2	45.6	47.0	50.6	76.2	49.7	78.2	82.1	0.3	96	244
MOT17-06-SDP	56.8	59.2	46.5	58	64	354	4,638	60.6	95.3	46.3	47.0	52.0	74.8	49.8	78.2	82.2	0.3	93	232
MOT17-07-DPM	40.5	42.5	33.5	6	24	363	9,603	43.2	95.3	35.5	32.2	38.3	73.4	33.7	74.4	79.4	0.7	90	267
MOT17-07-FRCNN	39.4	43.1	33.8	7	24	555	9,588	43.2	92.9	36.1	32.2	38.6	75.7	34.0	73.0	79.2	1.1	93	258
MOT17-07-SDP	41.2	42.6	33.9	7	20	596	9,231	45.4	92.8	34.9	33.6	37.3	75.9	35.5	72.6	79.0	1.2	111	331
MOT17-08-DPM	27.0	30.7	27.7	7	38	213	15,130	28.4	96.6	33.6	23.0	35.3	82.2	23.6	80.2	83.2	0.3	83	122
MOT17-08-FRCNN	27.1	31.7	28.1	9	38	197	15,119	28.4	96.8	34.7	22.9	36.9	81.1	23.5	79.9	83.1	0.3	74	108
MOT17-08-SDP	28.7	32.1	28.4	10	36	253	14,715	30.3	96.2	33.8	24.1	36.3	79.2	24.9	78.8	82.5	0.4	103	158
MOT17-12-DPM	45.6	55.2	44.2	15	44	88	4,596	47.0	97.9	52.2	37.5	56.1	78.7	38.8	80.9	83.2	0.1	29	57
MOT17-12-FRCNN	43.4	53.9	43.2	14	47	185	4,697	45.8	95.5	52.0	36.0	56.2	77.7	37.8	78.8	83.2	0.2	25	41
MOT17-12-SDP	45.3	56.9	44.7	17	44	212	4,492	48.2	95.2	53.0	37.9	56.5	79.5	39.7	78.5	82.9	0.2	34	60
MOT17-14-DPM	26.9	37.1	27.0	11	87	591	12,834	30.6	90.5	33.2	22.2	35.8	70.7	23.1	68.5	75.9	0.8	92	285
MOT17-14-FRCNN	27.1	38.1	28.4	12	79	1,202	12,139	34.3	84.1	33.5	24.6	36.9	66.6	26.2	64.1	75.0	1.6	132	396
MOT17-14-SDP	27.6	38.5	28.4	12	79	1,208	12,021	35.0	84.3	33.0	24.9	35.6	69.4	26.6	64.1	75.0	1.6	158	434

Raw data:

download

RTAT

Tracktor++

AFN17

Tracktor++: Tracktor++