Research map / 3D vision / Face and shape
Variational Autoencoders for Deforming 3D Mesh Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use latent dimension 128 rather than 16 (ablation) | Mesh VAE · Jumping held-out shapes | per-vertex reconstruction error: 4.9129 -> 4.4325 (×10^-4) | Helped |
| use latent dimension 128 rather than 16 (ablation) | Mesh VAE · Face held-out shapes | per-vertex reconstruction error: 8.3833 -> 7.8025 (×10^-2) | Helped |
| increase latent dimension from 128 to 256 (ablation) | Mesh VAE · Jumping held-out shapes | per-vertex reconstruction error: 4.4325 -> 4.2442 (×10^-4) | Helped |
| increase latent dimension from 128 to 256 (ablation) | Mesh VAE · Face held-out shapes | per-vertex reconstruction error: 7.8025 -> 7.9192 (×10^-2) | Hurt |
| use RIMD features instead of 3D coordinates or aligned coordinates (ablation) | Mesh VAE · tested held-out shape datasets | per-vertex reconstruction loss: 59%-94% compared with 3D coordinates; 27%-72% compared with aligned coordinates | Helped |
| vary mesh density or choose a different base mesh (ablation) | Mesh VAE · Chicken wings held-out shapes | per-vertex reconstruction error: 2.77; 2.49; 3.01; 2.85; 2.80 (×10^-4) | No clear effect |
| replace MSE reconstruction loss with cross-entropy loss and sigmoid output (ablation) | Mesh VAE · mesh feature reconstruction | training convergence: three times of epochs | Hurt |
| use the low-dimensional learned embedding rather than PCA, t-SNE, or NPE | shape embedding · Embedding retrieval (styles and poses) | AUC: Mesh VAE (0.5168) > t-SNE (0.4961) > PCA (0.2272) > NPE (0.1391) | Helped |
Closest papers by meaning
- Mesh-Based Autoencoders for Localized Deformation Component Analysis (AAAI 2018 · similarity 0.91)
- LaGeM: A Large Geometry Model for 3D Representation Learning and Diffusion (ICLR 2025 · similarity 0.87)
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (NeurIPS 2025 · similarity 0.87)
- Dora: Sampling and Benchmarking for 3D Shape Variational Auto-Encoders (CVPR 2025 · similarity 0.87)
- Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling (NeurIPS 2025 · similarity 0.87)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (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.