Research map / Probabilistic and scientific modeling / Molecules and proteins
Do Not Escape From the Manifold: Discovering the Local Coordinates on the Latent Space of GANs
Read the paperCheck 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 Local Basis directions instead of GANSpace or SeFa global directions | StyleGAN2 W-space traversal on FFHQ · FFHQ | FID and traversal image robustness | Helped |
| use Local Basis directions instead of GANSpace directions | StyleGAN2 semantic traversal · FFHQ | semantic factorization | Helped |
| use Local Basis directions instead of global directions | StyleGAN2 subspace traversal · StyleGAN2 W-space | traversal image stability | Helped |
| replace linear traversal with Iterative Curve-Traversal (ablation) | Local Basis on StyleGAN2 · StyleGAN2 W-space | traversal image stability | Helped |
| use stochastic step sizes during Iterative Curve-Traversal (ablation) | StyleCLIP-guided traversal of Old · StyleGAN2 FFHQ | diversity of image manipulation | Helped |
| choose each traversal direction by similarity to the global basis rather than the previous local direction (ablation) | Iterative Curve-Traversal guided by StyleCLIP · Old | target semantic consistency | Helped |
| follow the global direction only at the departure point, then track the previous local direction (ablation) | Iterative Curve-Traversal guided by StyleCLIP · Old | target semantic consistency | Hurt |
| use a global basis for linear traversal rather than manifold-tracing traversal | StyleCLIP traversal · Old | image quality and semantic variation | Hurt |
Closest papers by meaning
- Finding the Global Semantic Representation in GAN through Fréchet Mean (ICLR 2023 · similarity 0.91)
- Self-Supervised Enhancement of Latent Discovery in GANs (AAAI 2022 · similarity 0.89)
- A Geometric Analysis of Deep Generative Image Models and Its Applications (ICLR 2021 · similarity 0.89)
- Closed-Form Factorization of Latent Semantics in GANs (CVPR 2021 · similarity 0.89)
- Unsupervised Discovery of Interpretable Directions in the GAN Latent Space (ICML 2020 · similarity 0.89)
- Style Quantization for Data-Efficient GAN Training (CVPR 2025 · 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.