Research map / 3D vision / Face and shape
MeshAnything V2: Artist-Created Mesh Generation with Adjacent Mesh Tokenization
Yiwen Chen, Yikai Wang, Yihao Luo, Zhengyi Wang, Zilong Chen, Jun Zhu, Chi Zhang, Guosheng Lin
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 Adjacent Mesh Tokenization instead of naive three-vertex tokenization (ablation) | MeshAnything V2 · Objaverse eval | CD (x10^-2): 0.895 -> 0.874 | Improved |
| add AMT (ablation) | MeshAnything V2 · Objaverse eval | ECD (x10^-2): 4.832 -> 4.721 | Improved |
| add AMT (ablation) | MeshAnything V2 · Objaverse eval | NC: 0.924 -> 0.933 | Improved |
| add AMT (ablation) | MeshAnything V2 · Objaverse eval | S Ratio: 1.000 -> 0.497 | Improved |
| add AMT (ablation) | MeshAnything V2 · Objaverse eval | #V and #F: 302.4 -> 308.6 (#V), 556.7 -> 571.8 (#F) | Hurt |
| use AMT instead of baseline tokenization | small OPT-125M model · Tokenization comparison | CD (x10^-2): 2.478 -> 2.348 | Improved |
| use AMT instead of baseline tokenization | small OPT-125M model · Tokenization comparison | ECD (x10^-2): 18.21 -> 19.33 | Hurt |
| sort vertices and faces rather than leaving the stored order (ablation) | baseline tokenization · Tokenization comparison | CD (x10^-2): 8.151 -> 2.478 | Improved |
| use AMT instead of PolyGen(AMT) | mesh tokenization · Tokenization comparison | CD (x10^-2): 3.226 -> 2.348 | Improved |
| use PolyGen-style vertices-then-faces tokenization with AMT instead of AMT | mesh tokenization · Tokenization comparison | S Ratio: 0.492 -> 0.372 | Improved |
| add vertex swap special token to AMT (ablation) | AMT · Tokenization comparison | S Ratio: 0.492 -> 0.455 | Improved |
| add vertex swap special token to AMT (ablation) | AMT · Tokenization comparison | CD (x10^-2): 2.348 -> 2.517 | No clear effect |
Closest papers by meaning
- MeshAnything: Artist-Created Mesh Generation with Autoregressive Transformers (ICLR 2025 · similarity 0.91)
- Scaling Mesh Generation via Compressive Tokenization (CVPR 2025 · similarity 0.88)
- MeshGPT: Generating Triangle Meshes with Decoder-Only Transformers (CVPR 2024 · similarity 0.88)
- FreeMesh: Boosting Mesh Generation with Coordinates Merging (ICML 2025 · similarity 0.87)
- MeshArt: Generating Articulated Meshes with Structure-Guided Transformers (CVPR 2025 · similarity 0.87)
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (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.
