atlas

Research map / Image generation / Editing and style transfer

HyperStyle: StyleGAN Inversion With HyperNetworks for Real Image Editing

CVPR 2022 · Image generation › Editing and style transfer · approach: GAN inversion for editing · 259 citations (OpenAlex)

Check your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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 HyperStyle instead of the single-shot encoder ReStyle_pSpStyleGAN2 inversion · CelebA-HQ testID: 0.66 -> 0.76Helped
Use HyperStyle instead of PTI generator tuningStyleGAN2 inversion · CelebA-HQ testTime (s): 55.715 -> 1.234Helped
Add iterative refinement rather than using one pass (ablation)HyperStyle · CelebA-HQ testID: 0.66 -> 0.76Helped
Refine medium and fine layers only, excluding coarse layers (ablation)HyperStyle · CelebA-HQ testID: 0.74 -> 0.76No clear effect
Exclude toRGB layers from refinement (ablation)HyperStyle · CelebA-HQ testID: 0.67 -> 0.76Helped
Add Shared Refinement blocks (ablation)HyperStyle · CelebA-HQ testID: 0.68 -> 0.76Helped
Replace shared kernel offsets with separable-convolution offset predictions (ablation)HyperStyle · CelebA-HQ testID: 0.76 -> 0.71Hurt

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