Research map / 3D vision / NeRF and radiance fields
SinGRAF: Learning a 3D Generative Radiance Field for a Single Scene
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 |
|---|---|---|---|
| replace full-scale training with progressive patch scaling (ablation) | SinGRAF · frl_apt.4 | Div.: .001 -> .308 | Helped |
| add camera pose optimization (ablation) | SinGRAF with progressive patches · frl_apt.4 | KID: .046 -> .037 | Helped |
| add camera pose optimization (ablation) | SinGRAF with progressive patches · frl_apt.4 | Div.: .308 -> .295 | Hurt |
| add perspective augmentation (ablation) | SinGRAF with camera optimization · frl_apt.4 | Div.: .295 -> .335 | Helped |
| add perspective augmentation (ablation) | SinGRAF with camera optimization · frl_apt.4 | KID: .037 -> .037 | No clear effect |
| remove discriminator scale conditioning (ablation) | SinGRAF · frl_apt.4 | KID: .037 -> 0.070 | Hurt |
| train with 50 rather than 100 input images (ablation) | SinGRAF (128²) · frl_apt.4 | KID: .037 -> .071 | Hurt |
| train with 10 rather than 100 input images (ablation) | SinGRAF (128²) · frl_apt.4 | Div.: .335 -> .002 | Hurt |
| use SinGRAF instead of GSN | 3D scene generation · office_3 | Div.: .001 -> .297 | Helped |
| use SinGRAF instead of GSN | 3D scene generation · castle | KID: .050 -> .064 | Hurt |
Closest papers by meaning
- GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020 · similarity 0.91)
- Benchmarking and Analyzing 3D-aware Image Synthesis with a Modularized Codebase (NeurIPS 2023 · similarity 0.90)
- Pix2NeRF: Unsupervised Conditional p-GAN for Single Image to Neural Radiance Fields Translation (CVPR 2022 · similarity 0.90)
- 3D-Aware Image Synthesis via Learning Structural and Textural Representations (CVPR 2022 · similarity 0.90)
- EpiGRAF: Rethinking training of 3D GANs (NeurIPS 2022 · similarity 0.89)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 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.