Research map / 3D vision / Face and shape
MS3D: High-Quality 3D Generation via Multi-Scale Representation Modeling
Guan Luo, Jianfeng Zhang
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 multi-scale SAMv2 features instead of Dinov2 (ablation) | MS3D · GSO ablation (GT images input) | PSNR: 32.10 -> 34.08 | Improved |
| add multi-scale SAMv2 features instead of Dinov2 (ablation) | MS3D · GSO ablation (GT images input) | LPIPS-N: 0.050 -> 0.036 | Improved |
| add multi-scale features instead of single-scale SAMv2 (ablation) | MS3D · GSO ablation (GT images input) | PSNR: 30.16 -> 34.08 | Improved |
| add multi-scale features instead of single-scale SAMv2 (ablation) | MS3D · GSO ablation (GT images input) | Chamfer distance: 0.010 -> 0.010 | No clear effect |
| raise max voxel resolution from 64 to 256 (ablation) | MS3D · GSO ablation (GT images input) | PSNR-N: 30.56 -> 34.71 | Improved |
| raise max voxel resolution from 128 to 256 (ablation) | MS3D · GSO ablation (GT images input) | PSNR-N: 33.25 -> 34.71 | Improved |
| raise max voxel resolution from 64 to 256 (ablation) | MS3D · GSO ablation (GT images input) | F-Score: 0.981 -> 0.992 | Improved |
| add multi-scale rendering supervision (ablation) | MS3D · GSO ablation (GT images input) | PSNR: 32.32 -> 34.08 | Improved |
| add multi-scale rendering supervision (24h training comparison) (ablation) | MS3D · GSO ablation (GT images input) | PSNR: 32.32 -> 33.52 | Improved |
| replace MeshFormer with MS3D | MeshFormer · GSO single image to 3D | Chamfer distance: 0.031 -> 0.026 | Improved |
| replace TRELLIS with MS3D | TRELLIS · OmniObject3D single image to 3D | F-Score: 0.933 -> 0.963 | Improved |
| replace MeshFormer with MS3D | MeshFormer · OmniObject3D single image to 3D | PSNR: 18.85 -> 21.86 | 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.
- MeshFormer — MeshFormer : High-Quality Mesh Generation with 3D-Guided Reconstruction Model (NeurIPS 2024) · 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
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.88)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.88)
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · similarity 0.87)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.87)
- Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling (NeurIPS 2025 · similarity 0.87)
- Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image (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.
