Research map / 3D vision / NeRF and radiance fields
In-Place Scene Labelling and Understanding With Implicit Scene Representation
Shuaifeng Zhi, Tristan Laidlow, Stefan Leutenegger, Andrew J. Davison
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 |
|---|---|---|---|
| add multi-view consistent NeRF training to noisy pixel labels (90% flipped) | Semantic-NeRF · Replica pixel-wise denoising | mIoU: 0.041 -> 0.877 | Improved |
| add Semantic-NeRF training to region-wise noisy labels (Even, 30%) | Semantic-NeRF · Replica Room_2 region-wise denoising | mIoU: 0.741 -> 0.796 | Improved |
| add Semantic-NeRF training to region-wise noisy labels (Sort, 50%) | Semantic-NeRF · Replica Room_2 region-wise denoising | mIoU: 0.793 -> 0.803 | No clear effect |
| use sparse geometrically accurate low-res labels instead of interpolated dense ones (S=8) (ablation) | Semantic-NeRF label super-resolution · Replica label super-resolution | mIoU: 0.610 -> 0.887 | Improved |
| use sparse low-res labels instead of interpolated dense ones (S=16) (ablation) | Semantic-NeRF label super-resolution · Replica label super-resolution | mIoU: 0.433 -> 0.800 | Improved |
| increase labelled pixels per class from single click to 10% (ablation) | Semantic-NeRF label propagation · Replica label propagation | mIoU: 0.602 -> 0.884 | Improved |
| replace Bayesian fusion with NeRF-training fusion of CNN predictions | multi-view semantic fusion · Replica NYUv2-13 fusion | mIoU: 0.668 -> 0.680 | Improved |
| add NeRF-training fusion to monocular CNN predictions | DeepLabV3+ · Replica NYUv2-13 fusion | mIoU: 0.659 -> 0.680 | Improved |
| add average fusion to monocular CNN predictions | DeepLabV3+ · Replica NYUv2-13 fusion | mIoU: 0.659 -> 0.586 | Hurt |
| add semantic prediction head to NeRF (ablation) | NeRF · Replica reconstruction | PSNR: 32.80 -> 32.27 | No clear effect |
| add semantic prediction head to NeRF (ablation) | NeRF · Replica depth | RMSE: 0.102 -> 0.096 | No clear effect |
| set semantic loss weight lambda to 1 instead of 0.04 (ablation) | Semantic-NeRF · Replica | semantic performance | No clear effect |
Closest papers by meaning
- GP-NeRF: Generalized Perception NeRF for Context-Aware 3D Scene Understanding (CVPR 2024 · similarity 0.89)
- Enhancing Close-up Novel View Synthesis via Pseudo-labeling (AAAI 2025 · similarity 0.87)
- GSNeRF: Generalizable Semantic Neural Radiance Fields with Enhanced 3D Scene Understanding (CVPR 2024 · similarity 0.86)
- TextNeRF: A Novel Scene-Text Image Synthesis Method based on Neural Radiance Fields (CVPR 2024 · similarity 0.86)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.86)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 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.
