Research map / Image generation / Autoregressive and latent generation
Controlling generative models with continuous factors of variations
Antoine Plumerault, Hervé Le Borgne, Céline Hudelot
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use the Gaussian frequency-weighted reconstruction loss instead of MSE | Generative-model inversion · ILSVRC, 1000 categories | mean LPIPS: 0.57 -> 0.52 | Improved |
| Use the Gaussian frequency-weighted reconstruction loss instead of DSSIM | Generative-model inversion · ILSVRC, 1000 categories | mean LPIPS: 0.58 -> 0.52 | Improved |
| Use the Gaussian frequency-weighted reconstruction loss instead of MSE | Generative-model inversion · ILSVRC, 1000 categories | std LPIPS: 0.14 -> 0.12 | Improved |
| Use the Gaussian frequency-weighted reconstruction loss instead of DSSIM | Generative-model inversion · ILSVRC, 1000 categories | std LPIPS: 0.12 -> 0.12 | No clear effect |
| Increase β in the β-VAE (ablation) | β-VAE on dSprites · dSprites, object position | standard deviation of measured position over control parameter t | Improved |
| Constrain the optimized latent code to a ball of radius √d (ablation) | Generative-model inversion · ILSVRC reconstructions | reconstruction artifacts | Improved |
| Estimate transformed latent codes through sequential small transformations, initializing each step from the previous result | Latent-space trajectory estimation · Image transformations | optimization convergence | Improved |
| Learn a shared direction from merged category trajectories | BigGAN · ILSVRC validation categories, position and scale | control of transformation | Improved |
Closest papers by meaning
- Complexity Matters: Rethinking the Latent Space for Generative Modeling (NeurIPS 2023 · similarity 0.91)
- Navigating the GAN Parameter Space for Semantic Image Editing (CVPR 2021 · similarity 0.90)
- Unveiling the Latent Space Geometry of Push-Forward Generative Models (ICML 2023 · similarity 0.89)
- LARGE: Latent-Based Regression Through GAN Semantics (CVPR 2022 · similarity 0.89)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (NeurIPS 2024 · similarity 0.88)
- EQ-VAE: Equivariance Regularized Latent Space for Improved Generative Image Modeling (ICML 2025 · similarity 0.88)
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.
