CSTrack: Rethinking the competition between detection and ReID in Multi-Object Tracking


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:

Short name:

CSTrack

Detector:

Private

Description:

CSTrack proposes a strong ReID based one-shot MOT framework. It includes a novel cross-correlation network that can effectively impel the separate branches to learn task-dependent representations, and a scale-aware attention network that learns discriminative embeddings to improve the ReID capability. This work also provides an analysis of the weak data association ability in one-shot MOT methods. Our improvements make the data association ability of our one-shot model is comparable to two-stage methods while running more faster.

Reference:

C. Liang, Z. Zhang, Y. Lu, X. Zhou, B. Li. Rethinking the competition between detection and ReID in Multi-Object Tracking. In arXiv:2010.12138 [cs], 2020.

Last submitted:

November 17, 2020 (3 years ago)

Published:

October 21, 2020 at 15:50:12 CET

Submissions:

2

Project page / code:

Open source:

Yes

Hardware:

RTX2080ti

Runtime:

15.8 Hz

Benchmark performance:

Sequence MOTA IDF1 HOTA MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
MOT1774.972.659.3978 (41.5)411 (17.5)23,847114,30379.795.057.961.163.276.666.078.683.31.33,567 (44.7)7,668 (96.2)

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-DPM57.163.050.988892,64559.097.754.447.758.581.249.682.184.20.235103
MOT17-01-FRCNN57.163.050.988892,64559.097.754.447.758.581.249.682.184.20.235103
MOT17-01-SDP57.163.050.988892,64559.097.754.447.758.581.249.682.184.20.235103
MOT17-03-DPM89.282.167.813404,4416,61093.795.764.072.269.079.678.179.883.83.0263626
MOT17-03-FRCNN89.282.167.813404,4416,61093.795.764.072.269.079.678.179.883.83.0263626
MOT17-03-SDP89.282.167.813404,4416,61093.795.764.072.269.079.678.179.883.83.0263626
MOT17-06-DPM61.663.951.279544023,95966.495.151.351.258.973.054.778.483.20.3168286
MOT17-06-FRCNN61.663.951.279544023,95966.495.151.351.258.973.054.778.483.20.3168286
MOT17-06-SDP61.663.951.279544023,95966.495.151.351.258.973.054.778.483.20.3168286
MOT17-07-DPM64.961.248.72268804,90371.093.244.453.649.470.358.476.683.01.8152356
MOT17-07-FRCNN64.961.248.72268804,90371.093.244.453.649.470.358.476.683.01.8152356
MOT17-07-SDP64.961.248.72268804,90371.093.244.453.649.470.358.476.683.01.8152356
MOT17-08-DPM53.045.639.62599858,59059.392.736.044.743.059.048.375.582.11.6351548
MOT17-08-FRCNN53.045.639.62599858,59059.392.736.044.743.059.048.375.582.11.6351548
MOT17-08-SDP53.045.639.62599858,59059.392.736.044.743.059.048.375.582.11.6351548
MOT17-12-DPM56.566.653.926226173,10064.290.058.649.864.178.554.776.784.90.749176
MOT17-12-FRCNN56.566.653.926226173,10064.290.058.649.864.178.554.776.784.90.749176
MOT17-12-SDP56.566.653.926226173,10064.290.058.649.864.178.554.776.784.90.749176
MOT17-14-DPM51.361.242.932385358,29455.195.046.240.349.874.142.773.679.50.7171461
MOT17-14-FRCNN51.361.242.932385358,29455.195.046.240.349.874.142.773.679.50.7171461
MOT17-14-SDP51.361.242.932385358,29455.195.046.240.349.874.142.773.679.50.7171461

Raw data: