Research map / 3D vision / NeRF and radiance fields
GRAM: Generative Radiance Manifolds for 3D-Aware 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 |
|---|---|---|---|
| replace volume-wide hierarchical sampling with learned radiance manifolds (ablation) | GRAM · FFHQ 128² | FID 5K: 35.4 -> 25.8 | Helped |
| learn the surfaces rather than fixing parallel planes (ablation) | GRAM · FFHQ 128² | FID 5K: 28.3 -> 25.8 | Helped |
| learn the surfaces rather than fixing sphere-like initialized surfaces (ablation) | GRAM · FFHQ 128² | FID 5K: 27.8 -> 25.8 | Helped |
| use learned manifolds instead of hierarchical sampling at 6 samples (ablation) | GRAM · FFHQ 128² | FID 5K: 117 -> 27.4 | Helped |
| increase hierarchical sampling from 6 to 12 points (ablation) | NeRF-H · FFHQ 128² | FID 5K: 117 -> 62.6 | Helped |
| increase hierarchical sampling from 12 to 48 points (ablation) | NeRF-H · FFHQ 128² | FID 5K: 62.6 -> 30.0 | Helped |
| increase manifold surfaces from 6 to 24 (ablation) | GRAM · FFHQ 128² | FID 5K: 27.4 -> 25.8 | Helped |
| increase manifold surfaces from 24 to 36 (ablation) | GRAM · FFHQ 128² | FID 5K: 25.8 -> 25.8 | No clear effect |
| increase manifold surfaces from 24 to 48 (ablation) | GRAM · FFHQ 128² | FID 5K: 25.8 -> 25.2 | Helped |
| add real-pose regularization (ablation) | NeRF-H · FFHQ 128² | FID 5K: 44.4 -> 35.4 | Helped |
| add real-pose regularization (ablation) | GRAM · FFHQ 128² | FID 5K: 26.4 -> 25.8 | Helped |
| remove progressive growing (ablation) | GRAM · FFHQ 128² | FID 5K: 30.6 -> 28.8 | Helped |
Closest papers by meaning
- GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020 · similarity 0.91)
- Learning Detailed Radiance Manifolds for High-Fidelity and 3D-Consistent Portrait Synthesis From Monocular Image (CVPR 2023 · similarity 0.91)
- 3D-Aware Image Synthesis via Learning Structural and Textural Representations (CVPR 2022 · similarity 0.90)
- Benchmarking and Analyzing 3D-aware Image Synthesis with a Modularized Codebase (NeurIPS 2023 · similarity 0.90)
- SIGNeRF: Scene Integrated Generation for Neural Radiance Fields (CVPR 2024 · similarity 0.88)
- GaussianCube: A Structured and Explicit Radiance Representation for 3D Generative Modeling (NeurIPS 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.