Research map / 3D vision / Face and shape
MSVS-VAE: Multi-Scale Anchored VecSet for High-Fidelity 3D Reconstruction
Dehao Hao, Kaiyi Zhang, Tanghui Jia, Xiangjun Gao, Dongyu Yan, Weikai Chen, Zeyu HU, Lingting Zhu, Yingda Yin, Runze Zhang, Li Yuan, Xin Wang and 1 more
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 query decoding (ablation) | MSVS-VAE · Objaverse | Mesh Distance (x10^4): 1.690 -> 1.395 | Improved |
| add multi-scale query decoding (ablation) | MSVS-VAE · Objaverse | F1@0.001: 99.322 -> 99.621 | Improved |
| add multi-scale query decoding (ablation) | MSVS-VAE · in-the-wild | Mesh Distance (x10^4): 1.626 -> 1.231 | Improved |
| add hierarchical point-shuffle upsampling (ablation) | MSVS-VAE · Objaverse | Mesh Distance (x10^4): 3.693 -> 1.690 | Improved |
| add hierarchical point-shuffle upsampling (ablation) | MSVS-VAE · Objaverse | F1@0.001: 93.281 -> 99.322 | Improved |
| add hierarchical point-shuffle upsampling (ablation) | MSVS-VAE · in-the-wild | F1@0.001: 90.675 -> 99.416 | Improved |
| replace global cross-attention with AVS-Conv | MSVS-VAE vs Lattice · query latency, 40k tokens | latency (ms): 3068.05 -> 213.29 (14.4x) | Improved |
| increase token budget at inference (ablation) | MSVS-VAE · Objaverse | Mesh Distance (x10^4): 5.644 -> 1.395 (4k to 40k tokens) | Improved |
| use MSVS-VAE instead of Lattice at 4k tokens | MSVS-VAE · Objaverse | Mesh Distance (x10^4): 11.502 -> 5.644 | Improved |
| use MSVS-VAE 20k instead of SparseFlex-1024 (210k tokens) | MSVS-VAE · Objaverse | Mesh Distance (x10^4): 2.165 -> 1.695 | Improved |
| use MSVS-VAE-40K instead of SparseFlex on Dora-Bench Level-4 S-MD | MSVS-VAE-40K · Dora-Bench Level-4 | S-MD (x10^4): 3.08 -> 3.15 | 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.
- Lattice — LATTICE: Democratize High-Fidelity 3D Generation at Scale (CVPR 2026) · 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
- Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · similarity 0.88)
- WildFusion: Learning 3D-Aware Latent Diffusion Models in View Space (ICLR 2024 · similarity 0.87)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.87)
- LaGeM: A Large Geometry Model for 3D Representation Learning and Diffusion (ICLR 2025 · similarity 0.87)
- Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling (NeurIPS 2025 · similarity 0.87)
- Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention (NeurIPS 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.
