Research map / 3D vision / Gaussian splatting
G3R: Gradient Guided Generalizable Reconstruction
Yun Chen, Jingkang Wang, Ze Yang, Sivabalan Manivasagam, Raquel Urtasun
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 |
|---|---|---|---|
| replace 3D neural Gaussians with standard 3D Gaussian representation (ablation) | G3R · PandaSet | PSNR: 24.72 -> 25.22 (adding neural Gaussians) | Improved |
| add iterative reconstruction instead of one-step prediction (ablation) | G3R · PandaSet | PSNR: 20.03 -> 25.22 | Improved |
| add novel-view supervision across many scenes in training (ablation) | G3R · PandaSet | PSNR: 24.59 -> 25.22 | Improved |
| add decaying update schedule instead of constant scale (ablation) | G3R · PandaSet | PSNR: 25.03 -> 25.22 | Improved |
| use G3R instead of per-scene 3DGS optimization | scene reconstruction · PandaSet | reconstruction time: 50min 14s -> 123s with PSNR 25.14 -> 25.22 | Improved |
| use G3R instead of 3DGS on BlendedMVS | scene reconstruction · BlendedMVS | LPIPS: 0.462 -> 0.390 | Improved |
| use turbo variant (fewer iterations and Gaussians) | G3R · PandaSet | PSNR: 25.22 -> 24.76; 123s -> 31s | Mixed |
| adapt PixelSplat with 3D scaffold and all source images (PixelSplat++) | PixelSplat · BlendedMVS | PSNR: 16.24 -> 19.60 | Improved |
| finetune PandaSet-pretrained G3R on 2 BlendedMVS scenes (ablation) | G3R · BlendedMVS | PSNR: 24.11 -> 24.99 | 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.
- PixelSplat — pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024) · better on 3 of 3 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
- G3DR: Generative 3D Reconstruction in ImageNet (CVPR 2024 · similarity 0.90)
- GTR: Improving Large 3D Reconstruction Models through Geometry and Texture Refinement (ICLR 2025 · similarity 0.88)
- SaLon3R: Structure-Aware Long-Term Feedforward 3D Reconstruction from Unposed Images (IJCV 2026 · similarity 0.88)
- GenNBV: Generalizable Next-Best-View Policy for Active 3D Reconstruction (CVPR 2024 · similarity 0.87)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 2024 · similarity 0.87)
- MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data (CVPR 2025 · 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.
