AI Paper Atlas

Research map / Image generation / Editing and style transfer

Multi-Channel Attention Selection GANs for Guided Image-to-Image Translation

Hao Tang, Philip H. S. Torr, Nicu Sebe

TPAMI 2022 · Image generation › Editing and style transfer · approach: Image-to-image translation · 52 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 semantic guidance S_g alone instead of I_a (ablation)Pix2pix · Dayton a2g ablationSSIM: 0.4555 -> 0.5223Improved
add semantic guidance to input image (ablation)Pix2pix · Dayton a2g ablationSSIM: 0.4555 -> 0.5374Improved
add cycled semantic guidance generation (ablation)Baseline C · Dayton a2g ablationSSIM: 0.5374 -> 0.5438Improved
add uncertainty-guided pixel loss (ablation)Baseline D · Dayton a2g ablationSSIM: 0.5438 -> 0.5522Improved
add multi-channel attention selection (ablation)Baseline E · Dayton a2g ablationSSIM: 0.5522 -> 0.5989Improved
add total variation regularization (ablation)Baseline F · Dayton a2g ablationSSIM: 0.5989 -> 0.6047Improved
add multi-scale spatial pooling (ablation)Baseline G · Dayton a2g ablationSSIM: 0.6047 -> 0.6167Improved
add multi-scale spatial pooling (ablation)Baseline G · Dayton a2g ablationFID: 74.11 -> 72.23Improved
add second-stage refinement (coarse-to-fine) (ablation)Baseline H · Dayton a2g ablationSSIM: 0.5567 -> 0.6167Improved
use multi-scale spatial pooling instead of SE layer (ablation)SelectionGAN · Dayton a2g ablationSSIM: 0.5912 -> 0.6167Improved
use N=10 attention channels instead of N=1 (ablation)SelectionGAN · Dayton a2g ablationSSIM: 0.5522 -> 0.5986Improved
increase attention channels from 10 to 32 (ablation)SelectionGAN · Dayton a2g ablationSSIM: 0.5986 -> 0.5950No clear effect

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