AI Paper Atlas

Research map / Image generation / Image restoration

Jointly Aligning Millions of Images With Deep Penalised Reconstruction Congealing

Roberto Annunziata, Christos Sagonas, Jacques Cali

ICCV 2019 · Image generation › Image restoration · approach: Image harmonization and color transformation · 11 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 penalised low-capacity auto-encoder loss C to the reference-based similarity loss D (ablation)Proposed congealing on MNIST · MNIST (6000 images per digit)variance of aligned images (qualitative)Improved
add similarity loss D with reference image to C-only objective (ablation)Proposed congealing on MNIST · MNIST (6000 images per digit)alignment quality (qualitative)Improved
use proposed method instead of RASL on enlarged initial bounding box (+30%)LFW Jennifer_Capriati · LFW landmark errorlandmark error: 6.77% -> 1.92%Improved
coarser initialisation (+15% bounding box) for the proposed methodProposed on LFW George_W_Bush · LFW landmark errorlandmark error: 2.67% -> 3.84%Hurt
coarser initialisation (+30% bounding box) for RASLRASL on LFW George_W_Bush · LFW landmark errorlandmark error: 2.88% -> 6.29%Hurt
coarser initialisation (+30% bounding box) for the proposed methodProposed on LFW Tony_Blair · LFW landmark errorlandmark error: 2.39% -> 2.55%Mixed
use Deep Congealing instead of proposed method with original initialisationLFW George_W_Bush · LFW landmark errorlandmark error: 2.67% -> 3.97%Hurt

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