Research map / 3D vision / NeRF and radiance fields
DisCoScene: Spatially Disentangled Generative Radiance Fields for Controllable 3D-Aware Scene Synthesis
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 spatially disentangled object radiance fields with an abstract bounding-box layout prior | DisCoScene · Clevr | FID: 4.5 -> 3.5 | Helped |
| Use spatially disentangled object radiance fields with an abstract bounding-box layout prior | DisCoScene · 3D-Front | FID: 12.5 -> 13.8 | Hurt |
| Use spatially disentangled object radiance fields with an abstract bounding-box layout prior | DisCoScene · Waymo | FID: 15.1 -> 16.0 | Mixed |
| Add object-level discrimination (ablation) | DisCoScene · Clevr | FID: 5.0 -> 3.5 | Helped |
| Add object-level discrimination (ablation) | DisCoScene · 3D-Front | FID_obj: 33.7 -> 19.5 | Helped |
| Add object-level discrimination (ablation) | DisCoScene · Waymo | FID: 19.5 -> 16.0 | Helped |
| Add spatial conditioning from object location and scale (ablation) | DisCoScene · 3D-Front | FID: 15.2 -> 13.8 | Helped |
| Add spatial conditioning from object location and scale (ablation) | DisCoScene · 3D-Front | FID_obj: 23.2 -> 19.5 | Helped |
| Use efficient rendering that samples only valid points inside object boxes (ablation) | DisCoScene · Clevr | inference cost (ms/image): 180 -> 95 | Helped |
| Use efficient rendering that samples only valid points inside object boxes (ablation) | DisCoScene · 3D-Front | training cost (V100 days): 38.1 -> 22.3 | Helped |
| Use efficient rendering that samples only valid points inside object boxes (ablation) | DisCoScene · Waymo | inference cost (ms/image): 195 -> 100 | Helped |
| Use supersampling anti-aliasing during object synthesis (ablation) | DisCoScene · Object movement and boundary rendering | inference cost (ms/image): 95 -> 105 | Mixed |
Closest papers by meaning
- GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020 · similarity 0.91)
- SIGNeRF: Scene Integrated Generation for Neural Radiance Fields (CVPR 2024 · similarity 0.89)
- SinGRAF: Learning a 3D Generative Radiance Field for a Single Scene (CVPR 2023 · similarity 0.89)
- Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis (CVPR 2020 · similarity 0.88)
- Video Perception Models for 3D Scene Synthesis (NeurIPS 2025 · similarity 0.86)
- BerfScene: Bev-conditioned Equivariant Radiance Fields for Infinite 3D Scene Generation (CVPR 2024 · similarity 0.86)
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.