Research map / Probabilistic and scientific modeling / Variational and generative sampling
GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations
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 an autoregressive prior that conditions each scene component on previously generated components | GENESIS · GQN scene generation | coherent component-wise scene generation | Helped |
| Use an autoregressive prior that models relationships between components | GENESIS · ShapeStacks test set | ARI: 0.63±0.07 -> 0.73±0.03 | Helped |
| Use an autoregressive prior that models relationships between components | GENESIS · ShapeStacks test set | SC: 0.52±0.09 -> 0.64±0.08 | Helped |
| Use an autoregressive prior that models relationships between components | GENESIS · ShapeStacks test set | mSC: 0.49±0.09 -> 0.60±0.09 | Helped |
| Use object-centric latent representations | GENESIS · ShapeStacks stability | accuracy: 59.6 -> 64.0 | Helped |
| Use object-centric latent representations | GENESIS · ShapeStacks height | accuracy: 67.5 -> 80.3; 88.4 -> 80.3 | Mixed |
| Use a single latent per component rather than separate mask and appearance latents (ablation) | GENESIS · ShapeStacks stability | accuracy: 64.0 -> 63.2 | Hurt |
| Use a single latent per component rather than separate mask and appearance latents (ablation) | GENESIS · ShapeStacks height | accuracy: 80.3 -> 80.8 | Helped |
| Use a single latent per component rather than separate mask and appearance latents (ablation) | GENESIS · ShapeStacks view | accuracy: 99.3 -> 99.7 | Mixed |
| Use a single latent per component rather than separate mask and appearance latents (ablation) | GENESIS · Multi-dSprites | FID: 24.9 -> 28.2 | Hurt |
| Use a single latent per component rather than separate mask and appearance latents (ablation) | GENESIS · GQN | FID: 80.5 -> 70.2 | Helped |
| Use separate latent variables for masks and component appearances (ablation) | GENESIS · GQN scene generation | component interpretability | Helped |
Closest papers by meaning
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.91)
- DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos (NeurIPS 2024 · similarity 0.88)
- LT3SD: Latent Trees for 3D Scene Diffusion (CVPR 2025 · similarity 0.88)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.88)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.87)
- AutoPartGen: Autoregressive 3D Part Generation and Discovery (NeurIPS 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.