Research map / 3D vision / Face and shape
MagnetGS-Mesh: High-Quality Multi-Object Mesh Reconstruction via Adaptive Surface Optimization
Min-Su Park, Yeonho Han, Uijoon Jeong, Jun-Hyeong Park, Eun-Seok Ryu
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 Stage 1 enhanced occupancy losses to MILo (ablation) | MILo · LERF mesh rendering | PSNR: 18.93 -> 19.24 | Improved |
| add L_boundary to Stage 1 losses (ablation) | MILo · LERF mesh rendering | PSNR: 19.32 -> 19.48 | Improved |
| add L_smooth to Stage 1 losses (ablation) | MILo · LERF mesh rendering | SSIM: 0.6950 -> 0.7120 | Improved |
| add segmentation (SAGA) after Stage 1 (ablation) | MagnetGS-Mesh · LERF mesh rendering | PSNR: 19.24 -> 19.61 | Improved |
| add ALNC loss after segmentation (ablation) | MagnetGS-Mesh · LERF mesh rendering | PSNR: 19.61 -> 19.85 | Improved |
| add ALNC loss after segmentation (ablation) | MagnetGS-Mesh · LERF mesh storage | mesh size (MB): 14.2 -> 12.4 | Improved |
| add ALNC loss after segmentation (ablation) | MagnetGS-Mesh · LERF mesh rendering | LPIPS: 0.3040 -> 0.3088 | Hurt |
| average normals over neighbors instead of a single normal (ablation) | ALNC · LERF Stage 2 | PSNR: 19.68 -> 19.85 | Improved |
| add normal consistency to proximity loss (ablation) | ALNC · LERF Stage 2 | PSNR: 19.70 -> 19.85 | Improved |
| use density-adaptive k instead of fixed k = 10 (ablation) | ALNC · LERF Stage 2 | PSNR: 19.76 -> 19.85 | Improved |
| vary k range (kmin, kmax) around default 3 and 15 (ablation) | ALNC · LERF Stage 2 | PSNR: 19.74 - 19.85 | No clear effect |
| apply cell culling to the mesh (ablation) | MagnetGS-Mesh · DTU | Chamfer distance: 0.66 -> 0.71 | Hurt |
Closest papers by meaning
- Evolving High-Quality Rendering and Reconstruction in a Unified Framework with Contribution-Adaptive Regularization (CVPR 2025 · similarity 0.88)
- Dynamic Gaussians Mesh: Consistent Mesh Reconstruction from Dynamic Scenes (ICLR 2025 · similarity 0.88)
- Sparse2DGS: Geometry-Prioritized Gaussian Splatting for Surface Reconstruction from Sparse Views (CVPR 2025 · similarity 0.87)
- FatesGS: Fast and Accurate Sparse-View Surface Reconstruction Using Gaussian Splatting with Depth-Feature Consistency (AAAI 2025 · similarity 0.87)
- GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction (NeurIPS 2025 · similarity 0.87)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (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.
