AI Paper Atlas

Research map / Image generation / Image restoration

Context-Aware Image Matting for Simultaneous Foreground and Alpha Estimation

Qiqi Hou, Feng Liu

ICCV 2019 · Image generation › Image restoration · approach: Image harmonization and color transformation · 174 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
add context encoder (ablation)ME + L_lap · Composition-1K (alpha)SAD: 43.9 -> 35.8Improved
make matting-encoder-only model deeper (117.0M params) instead of adding context encoder (ablation)ME (model 2) · Composition-1K (alpha)SAD: 43.9 -> 43.7No clear effect
replace Deep Matting loss with Laplacian loss (ablation)ME · Composition-1K (alpha)SAD: 49.1 -> 43.9Improved
add feature loss (ablation)ME + CE + L_lap · Composition-1K (alpha)SAD: 35.8 -> 38.8Hurt
add feature loss (ablation)ME + CE + L_lap · Real-world images (NIMA score)NIMA mean score: 4.64 -> 4.69No clear effect
add foreground decoder color losses (L1c, LFc) (ablation)ME + CE + L_lap + L_F · Composition-1K (alpha)SAD: 38.8 -> 38.0Improved
add foreground decoder color losses (L1c, LFc) with DA (ablation)ME + CE + L_lap + L_F + DA · Composition-1K (alpha)SAD: 71.3 -> 84.1Hurt
add data augmentation (ablation)ME + CE + L_lap + L_F · Composition-1K (alpha)SAD: 38.8 -> 71.3Hurt
add data augmentation (ablation)ME + CE + L_lap + L_F · Real-world images (NIMA score)NIMA mean score: 4.69 -> 5.03Improved
add ReJPEGing augmentation (ablation)model 7 without ReJPEGing · Composition-1K (alpha)SAD: 55.1 -> 84.1Hurt
add Gaussian blur augmentation (ablation)model 7 without GaussianBlur · Composition-1K (alpha)SAD: 69.1 -> 84.1Hurt
add data augmentation (ablation)ME + CE + L_lap + L_F + color losses · Composition-1K (foreground)SAD: 61.72 -> 94.41Hurt

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