PermaTrack: Learning to Track with Object Permanence


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:

MOT17 |

Short name:

PermaTrack

Detector:

Public

Description:

Reference:

P. Tokmakov, J. Li, W. Burgard, A. Gaidon. Learning to Track with Object Permanence. In ICCV, 2021.

Last submitted:

May 07, 2021 (3 years ago)

Published:

May 07, 2021 at 04:53:20 CET

Submissions:

4

Project page / code:

n/a

Open source:

No

Hardware:

1 Tesla V100

Runtime:

11.9 Hz

Benchmark performance:

Sequence MOTA IDF1 HOTA MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
MOT1773.167.254.2996 (42.3)450 (19.1)24,577123,50878.194.751.258.059.667.962.876.181.61.43,571 (0.0)5,826 (0.0)

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-DPM50.448.138.91087122,46161.884.936.042.344.257.848.466.580.21.62750
MOT17-01-FRCNN49.848.539.01187892,41862.583.636.342.043.759.548.765.280.11.83056
MOT17-01-SDP50.748.339.21288242,32164.083.436.042.843.359.649.965.079.91.83260
MOT17-03-DPM87.878.962.312851,93610,64589.898.057.667.866.771.371.978.481.41.3164395
MOT17-03-FRCNN88.180.263.012851,91310,36590.198.058.768.066.972.472.178.481.41.3161377
MOT17-03-SDP88.580.363.512842,0219,83090.697.959.468.366.973.672.578.381.31.3150377
MOT17-06-DPM60.145.638.180525913,92266.793.029.949.450.343.153.574.681.20.5194279
MOT17-06-FRCNN61.047.539.282456843,69568.692.231.150.149.944.654.873.781.10.6216298
MOT17-06-SDP60.847.239.182476763,73868.392.231.050.049.944.554.673.881.10.6202290
MOT17-07-DPM59.848.640.82242,0414,52073.285.834.449.840.158.757.867.780.14.1228368
MOT17-07-FRCNN59.449.241.32541,9934,63872.586.035.449.440.761.157.267.880.14.0231367
MOT17-07-SDP60.047.940.82542,1034,42173.885.634.150.139.759.358.367.580.14.2226373
MOT17-08-DPM50.445.539.624158069,49955.093.537.342.844.159.945.877.883.61.3180251
MOT17-08-FRCNN49.045.539.623157859,80753.693.538.141.845.061.244.778.083.81.3178259
MOT17-08-SDP50.846.039.724158489,36555.793.337.243.244.060.846.377.683.51.4186266
MOT17-12-DPM50.658.149.827301,1203,11764.083.252.747.162.570.554.871.285.61.24174
MOT17-12-FRCNN54.459.750.727327213,19063.288.453.048.662.970.254.175.785.70.83867
MOT17-12-SDP52.659.050.327299683,09864.385.252.748.162.769.955.072.985.51.14177
MOT17-14-DPM52.051.940.135398477,69758.492.737.343.442.861.346.874.481.41.1321491
MOT17-14-FRCNN51.551.540.037411,1257,48559.590.736.943.742.460.847.872.881.31.5362517
MOT17-14-SDP52.952.440.939401,0747,27660.691.337.744.643.161.548.773.281.31.4363534

Raw data: