Research map / Image generation / GANs
Navigating the GAN Parameter Space for Semantic Image Editing
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 |
|---|---|---|---|
| Navigate generator parameters instead of latent codes | StyleGAN2 · LSUN-Cars, wheel size | ability to reproduce the visual effect | Helped |
| Use SVD-based parameter shifts instead of optimization-based, spectrum-based, or hybrid directions | StyleGAN2 · LSUN-Cars and LSUN-Horse | visual realism | Hurt |
| Use optimization-based, spectrum-based, or hybrid directions instead of SVD shifts | StyleGAN2 · LSUN-Cars, wheel size | FID | Helped |
| Choose one of the three proposed direction-discovery methods over the others | StyleGAN2 · LSUN-Cars, wheel size | FID | No clear effect |
| Use the hybrid scheme rather than SVD, optimization-based, or spectrum-based discovery | StyleGAN2 · LSUN-Horse, fourth layer | number and visual quality of discovered controls | Helped |
| Shift generator parameters rather than latent codes | StyleGAN2 · LSUN-Cars, wheel size | reproduction of the manipulation | Helped |
| Apply discovered parameter directions to inverted real images in W+ | StyleGAN2 · Real FFHQ faces and car images | applicability of semantic editing | Helped |
| Apply parameter shifts at different generator depths | StyleGAN2 · LSUN-Horse | type of visual effect | Mixed |
Closest papers by meaning
- Linear Semantics in Generative Adversarial Networks (CVPR 2021 · similarity 0.93)
- Editing in Style: Uncovering the Local Semantics of GANs (CVPR 2020 · similarity 0.93)
- EditGAN: High-Precision Semantic Image Editing (NeurIPS 2021 · similarity 0.92)
- Exploiting Spatial Dimensions of Latent in GAN for Real-Time Image Editing (CVPR 2021 · similarity 0.92)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.91)
- Reference-Based 3D-Aware Image Editing with Triplanes (CVPR 2025 · similarity 0.89)
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.