AI Paper Atlas

Research map / Video / Video segmentation and tracking

One-Trimap Video Matting

Hongje Seong, Seoung Wug Oh, Brian Price, Euntai Kim, Joon-Young Lee

ECCV 2022 · Video › Video segmentation and tracking · approach: Video restoration with temporal priors · 18 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
train trimap propagation and alpha prediction jointly rather than decoupled (ablation)STM+FBA · VideoMatting108 val (1-trimap)SSDA-V: 83.61 -> 75.36Improved
add stage-wise training on top of joint training (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 72.07 -> 54.67Improved
add pretraining stage 1 before end-to-end stage (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 87.31 -> 76.55Improved
add pretraining stage 2 (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 76.55 -> 75.33Improved
add pretraining stage 3 (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 75.33 -> 54.67Improved
add alpha and trimap refinement outputs (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 83.56 -> 77.58Improved
feed alpha matte to trimap module (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 77.58 -> 62.47Improved
feed hidden features of refinement module to trimap module (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 62.47 -> 54.67Improved
use OTVM with DIM backbone instead of STM+DIMOTVM · VideoMatting108 val (DIM backbone)SSDA-V: 102.77 -> 92.30Improved
use OTVM with GCA backbone instead of STM+GCAOTVM · VideoMatting108 val (GCA backbone)SSDA-V: 100.22 -> 84.73Improved
add explicit trimap input to trimap encoder (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 75.08 -> 54.67Improved
remove trimap input and alpha, keep only hidden for trimap encoder (ablation)OTVM · VideoMatting108 val (1-trimap)SSDA-V: 54.67 -> 130.54Hurt

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