AI Paper Atlas

Research map / Video / Video segmentation and tracking

End-to-End Video Matting With Trimap Propagation

Wei-Lun Huang, Ming-Sui Lee

CVPR 2023 · Video › Video segmentation and tracking · approach: Video semantic and panoptic segmentation · 14 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use FTP-VM rather than the strongest reported baselinevideo matting with trimap propagation · VM108MSE: 2.92 -> 5.19Hurt
Use FTP-VM rather than the strongest reported baselinevideo matting with trimap propagation · VM240kMSE: 0.56 -> 0.45Improved
Use FTP-VM rather than the strongest reported baselinevideo matting with trimap propagation · Real HumanMSE: 2.38 -> 2.80Hurt
Combine trimap propagation and matting in one network rather than use a two-stage network (ablation)FTP-VM · VM108MSE: 8.80 -> 10.21Hurt
Use separate, specially designed segmentation and matting decoders instead of the same decoders (ablation)FTP-VM · VM108MSE: 9.72 -> 5.19Improved
Replace an additional backbone with vanilla convolutions in trimap fusion (ablation)FTP-VM · VM108MSE: 11.34 -> 8.12Improved
Replace vanilla convolutions with gated convolutions in trimap fusion (ablation)FTP-VM · VM108MSE: 8.12 -> 5.19Improved
Add segmentation consistency loss (ablation)FTP-VM trimap segmentation · VM108MSE: 11.66 -> 5.19Improved
Apply consistency loss to all classes rather than only the correct class (ablation)FTP-VM trimap segmentation · VM108MSE: 6.19 -> 8.75Hurt
Weight consistency loss differently for correct and incorrect classes instead of applying it unweighted to all classes (ablation)FTP-VM trimap segmentation · VM108MSE: 8.75 -> 5.19Improved
Use weighted consistency loss across all classes rather than correct-class-only consistency (ablation)FTP-VM trimap segmentation · VM108dtSSD: 2.28 -> 2.46Hurt
Use the end-to-end FTP-VM instead of the two-model approachestrimap propagation and video matting · inferenceFPS: 5.11 -> 40.16Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy