Research map / 3D vision / Gaussian splatting
MGSR: 2D/3D Mutual-boosted Gaussian Splatting for High-fidelity Surface Reconstruction under Various Light Conditions
Qingyuan Zhou, Yuehu Gong, Weidong Yang, Jiaze Li, Yeqi Luo, Baixin Xu, Shuhao Li, Ben Fei, Ying He
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 mutual-boosted 2D/3D optimization versus 3DGS branch alone (ablation) | MGSR · OmniObject3D | PSNR: 35.10 -> 37.69 | Improved |
| add mutual-boosted optimization versus 2DGS branch alone (ablation) | MGSR · OmniObject3D | Normal Consistency: 90.16 -> 90.60 | Improved |
| use unidirectional rather than bidirectional back-propagation (ablation) | MGSR · OmniObject3D | PSNR: 36.64 -> 37.69 | Improved |
| add auto-stop warm-up strategy (ablation) | MGSR · OmniObject3D | PSNR: 37.01 -> 37.69 | Improved |
| increase mutual-boosted iterations from 20k to 30k (ablation) | MGSR · OmniObject3D | PSNR: 37.69 -> 37.71 | No clear effect |
| reduce mutual-boosted iterations from 20k to 5k (ablation) | MGSR · OmniObject3D | PSNR: 37.69 -> 36.76 | Hurt |
| raise depth mutual loss weight from 0.01 to 0.1 (ablation) | MGSR · OmniObject3D | PSNR: 37.69 -> 36.42 | Hurt |
| weight 3DGS branch higher (0.3/0.7) instead of equal weights (ablation) | MGSR · OmniObject3D | PSNR: 37.69 -> 37.88 | Mixed |
| replace 3D-GS/R3DG-style BRDF decomposition with geometry-guided decomposition | MGSR · OmniObject3D | training time: R3DG 21min -> MGSR 13min | Improved |
| compare MGSR against NeuS2 on reconstruction | MGSR · Shiny Blender / OmniObject3D | Chamfer Distance: NeuS2 0.48 vs MGSR 0.90 (OmniObject3D) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- NeuS2 — NeuS2: Fast Learning of Neural Implicit Surfaces for Multi-view Reconstruction (ICCV 2023) · better on 0 of 2 shared results
- GS-IR — GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024) · better on 1 of 1 shared result
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
- GS-2DGS: Geometrically Supervised 2DGS for Reflective Object Reconstruction (CVPR 2025 · similarity 0.93)
- Anti-Aliased 2D Gaussian Splatting (NeurIPS 2025 · similarity 0.92)
- VA-GS: Enhancing the Geometric Representation of Gaussian Splatting via View Alignment (NeurIPS 2025 · similarity 0.91)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.91)
- Normal-GS: 3D Gaussian Splatting with Normal-Involved Rendering (NeurIPS 2024 · similarity 0.91)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.91)
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.
