Research map / 3D vision / NeRF and radiance fields
Learning Object-Compositional Neural Radiance Field for Editable Scene Rendering
Bangbang Yang, Yinda Zhang, Yinghao Xu, Yijin Li, Han Zhou, Hujun Bao, Guofeng Zhang, Zhaopeng Cui
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 object-compositional NeRF scene branch instead of plain NeRF | Our scene branch · ScanNet (full scene rendering) | PSNR: 28.927 -> 29.005 | No clear effect |
| use object-compositional NeRF scene branch instead of plain NeRF | Our scene branch · ToyDesk (full scene rendering) | LPIPS: 0.537 -> 0.522 | Improved |
| use our scene branch instead of Sparse Voxel | Our scene branch · ScanNet (full scene rendering) | PSNR: 24.143 -> 29.005 | Improved |
| add scene guidance and 3D guard mask (ablation) | Object branch · ScanNet 0033 | PSNR: 19.785 -> 22.600 | Improved |
| add scene guidance and 3D guard mask (ablation) | Object branch · ScanNet 0038 | PSNR: 27.100 -> 34.435 | Improved |
| add scene guidance on top of 3D guard mask (ablation) | Object branch · ScanNet 0033 | PSNR: 20.450 -> 22.600 | Improved |
| add 3D guard mask on top of scene guidance (ablation) | Object branch · ScanNet 0033 | LPIPS: 0.117 -> 0.049 | Improved |
| add object voxel feature f_obj at object branch input (ablation) | Object branch · ScanNet 0033 | PSNR: 22.219 -> 22.600 | Improved |
| add object voxel feature f_obj at object branch input (ablation) | Object branch · ScanNet 0038 | PSNR: 33.861 -> 34.435 | Improved |
| vary 3D guard mask forward distance epsilon from 0.05 to 0.1 (ablation) | 3D guard mask · ScanNet 0033 | PSNR: 22.600 -> 22.172 | Hurt |
| vary 3D guard mask forward distance epsilon from 0.05 to 0.025 (ablation) | 3D guard mask · ScanNet 0033 | PSNR: 22.600 -> 22.364 | Hurt |
| render objects from object branch instead of clipping scene-branch rays to mesh radius or bounding box | Object branch · ScanNet | qualitative object rendering | Improved |
Closest papers by meaning
- Learning Unified Decompositional and Compositional NeRF for Editable Novel View Synthesis (ICCV 2023 · similarity 0.93)
- Structured-NeRF: Hierarchical Scene Graph with Neural Representation (ECCV 2024 · similarity 0.91)
- Neural Radiance Field Codebooks (ICLR 2023 · similarity 0.90)
- Neural Relightable Participating Media Rendering (NeurIPS 2021 · similarity 0.89)
- SIGNeRF: Scene Integrated Generation for Neural Radiance Fields (CVPR 2024 · similarity 0.88)
- Neural Directional Encoding for Efficient and Accurate View-Dependent Appearance Modeling (CVPR 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.
