Research map / 3D vision / Gaussian splatting
RetrievalFuse: Neural 3D Scene Reconstruction With a Database
Yawar Siddiqui, Justus Thies, Fangchang Ma, Qi Shan, Matthias Nießner, Angela Dai
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 attention blending instead of naive feature concatenation (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.71 -> 0.75 | Improved |
| add attention blending instead of naive feature concatenation (ablation) | RetrievalFuse · 3DFront surface reconstruction | IoU: 0.84 -> 0.86 | Improved |
| add retrievals to refinement compared with a U-Net without retrievals (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.68 -> 0.75 | Improved |
| add retrievals to refinement compared with a U-Net without retrievals (ablation) | RetrievalFuse · 3DFront surface reconstruction | IoU: 0.83 -> 0.86 | Improved |
| add refinement on top of 1-NN retrieval (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.67 -> 0.75 | Improved |
| increase retrieved chunks k from 0 to 4 (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.684 -> 0.751 | Improved |
| use smaller retrieval chunks (3.467m to 0.867m) (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.71 -> 0.75 | Improved |
| increase database train scenes from 25% to 100% (ablation) | RetrievalFuse · 3DFront 3D super-resolution | IoU: 0.711 -> 0.751 | Improved |
| extend database at test time with chunks of new classes (ablation) | RetrievalFuse · ShapeNet 8 -> 5 unseen classes | IoU: 0.478 -> 0.579 | Improved |
| add retrieval-based reconstruction to IFNet | IFNet · 3DFront 3D super-resolution | IoU: 0.639 -> 0.687 | Improved |
| use attention instead of naive fusion when retrievals are suboptimal (ablation) | RetrievalFuse · ShapeNet 8 -> 5 unseen classes | IoU: 0.432 -> 0.478; backbone 0.463 | Mixed |
| RetrievalFuse vs IFNet on Matterport3D surface reconstruction | RetrievalFuse · Matterport3D surface reconstruction | IoU: 0.575 -> 0.710 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SGNN — SG-NN: Sparse Generative Neural Networks for Self-Supervised Scene Completion of RGB-D Scans (CVPR 2020) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.90)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · similarity 0.88)
- MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data (CVPR 2025 · similarity 0.88)
- Denoising Diffusion via Image-Based Rendering (ICLR 2024 · similarity 0.88)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 2024 · similarity 0.87)
- SCube: Instant Large-Scale Scene Reconstruction using VoxSplats (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.
