Research map / Image generation / Editing and style transfer
StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use StyleSpace instead of Z | StyleGAN2 latent representation · FFHQ | DCI disentanglement: 0.31 -> 0.75 | Helped |
| Use StyleSpace instead of W | StyleGAN2 latent representation · FFHQ | DCI completeness: 0.57 -> 0.87 | Helped |
| Use StyleSpace instead of W | StyleGAN2 latent representation · FFHQ | DCI informativeness: 0.97 -> 0.99 | Helped |
| Use StyleSpace instead of W+ | StyleGAN2 latent representation · FFHQ, W+ comparison | DCI disentanglement: 0.54 -> 0.63 | Helped |
| Use StyleSpace instead of W+ | StyleGAN2 latent representation · FFHQ, W+ comparison | DCI completeness: 0.64 -> 0.81 | Helped |
| Use StyleSpace instead of W+ | StyleGAN2 latent representation · FFHQ, W+ comparison | DCI informativeness: 0.94 -> 0.98 | Helped |
| Restrict attribute-channel search to locally-active channels in the target region | StyleSpace exemplar-based attribute detection · FFHQ; sideburns, smile, and gray hair | top-5 detection rate: exceeds 92% using as few as 20 examples | Helped |
| Optimize inversion in S rather than W+ or W | StyleGAN2 real-image inversion · FFHQ real-image reconstruction | visual reconstruction accuracy | Helped |
| Optimize inversion in S rather than W or W+ | StyleGAN2 real-image inversion · FFHQ real-image manipulation | manipulation naturalness | Hurt |
| Initialize S-space latent optimization with an encoder embedding and fine-tune for a small number of iterations | StyleGAN2 real-image inversion · FFHQ and CelebA-HQ real images | artifact-free manipulation | Helped |
| Use StyleSpace controls instead of GANSpace or InterFaceGAN directions | StyleGAN2 attribute manipulation · FFHQ; Gender, Gray hair, and Lipstick | mean-AD | Helped |
Closest papers by meaning
- A Style is Worth One Code: Unlocking Code-to-Style Image Generation with Discrete Style Space (CVPR 2026 · similarity 0.89)
- SemanticStyleGAN: Learning Compositional Generative Priors for Controllable Image Synthesis and Editing (CVPR 2022 · similarity 0.89)
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 2025 · similarity 0.89)
- The Devil is in the Details: StyleFeatureEditor for Detail-Rich StyleGAN Inversion and High Quality Image Editing (CVPR 2024 · similarity 0.87)
- CSGO: Content-Style Composition in Text-to-Image Generation (NeurIPS 2025 · similarity 0.87)
- SDGAN: Disentangling Semantic Manipulation for Facial Attribute Editing (AAAI 2024 · similarity 0.87)
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.