Research map / 3D vision / Gaussian splatting
Generalizable Patch-Based Neural Rendering
Mohammed Suhail, Carlos Esteves, Leonid Sigal, Ameesh Makadia
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 |
|---|---|---|---|
| fine-tune the model 10k iterations per scene and compare to IBRNet | Ours · RFF (fine-tuned, setting 1) | PSNR: 26.73 -> 27.66 | Improved |
| fine-tune the model vs GeoNeRF | Ours · RFF (fine-tuned, setting 1) | LPIPS: 0.162 -> 0.138 | Improved |
| increase reference views from 3 to 5 (ablation) | Ours · Real-Forward-Facing | PSNR: 22.36 -> 24.33 | Improved |
| increase reference views from 5 to 7 (ablation) | Ours · Real-Forward-Facing | PSNR: 24.33 -> 25.05 | Improved |
| increase reference views from 7 to 10 (ablation) | Ours · Real-Forward-Facing | PSNR: 25.05 -> 25.72 | Improved |
| increase reference views from 10 to 12 (ablation) | Ours · Real-Forward-Facing | PSNR: 25.72 -> 25.69 | No clear effect |
| combine reference pixel colors instead of aggregated features (ablation) | Ours · Real-Forward-Facing (setting 1) | PSNR: 25.08 -> 25.72 | Improved |
| combine colors instead of features (ablation) | Ours · Real-Forward-Facing (setting 1) | LPIPS: 0.199 -> 0.175 | Improved |
| compare to MVSNeRF when trained only on DTU | Ours · RFF (trained on DTU) | PSNR: 21.93 -> 20.69 | Hurt |
| compare to MVSNeRF when trained only on DTU | Ours · RFF (trained on DTU) | SSIM: 0.795 -> 0.808 | Improved |
| use this model instead of per-scene LFNR for rendering | Ours · Timing | rendering time per image: 10 -> 15 | Hurt |
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.
- GeoNeRF — GeoNeRF: Generalizing NeRF With Geometry Priors (CVPR 2022) · better on 2 of 2 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
- DynIBaR: Neural Dynamic Image-Based Rendering (CVPR 2023 · similarity 0.86)
- Geometry-aware Reconstruction and Fusion-refined Rendering for Generalizable Neural Radiance Fields (CVPR 2024 · similarity 0.85)
- GAURA: Generalizable Approach for Unified Restoration and Rendering of Arbitrary Views (ECCV 2024 · similarity 0.83)
- Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation (ICLR 2024 · similarity 0.82)
- RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations (NeurIPS 2026 · similarity 0.82)
- Neural Directional Encoding for Efficient and Accurate View-Dependent Appearance Modeling (CVPR 2024 · similarity 0.81)
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.
