Research map / 3D vision / Face and shape
Leveraging 2D Data to Learn Textured 3D Mesh Generation
Check 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 |
|---|---|---|---|
| Train without segmentation-mask annotations | Textured 3D mesh model · ShapeNet classes | IS: IS: airplane 4.0 -> 3.2; car 4.1 -> 3.5; chair 5.8 -> 5.2; sofa 4.3 -> 4.3 | Hurt |
| Train without segmentation-mask annotations | Textured 3D mesh model · ShapeNet classes | FID: airplane 73.5 -> 56.5; car 154.0 -> 165.4; chair 111.1 -> 82.6; sofa 58.3 -> 63.8 | Mixed |
| Train without segmentation-mask annotations | Textured 3D mesh model · Natural-image birds | FID: 104.3 -> 75.4 | Helped |
| Train without segmentation-mask annotations | Textured 3D mesh model · Natural-image cars | FID: 191.4 -> 182.1 | Helped |
| Replace dense vertex decoder with pushing parametrization (ablation) | Textured 3D mesh model · ShapeNet airplane, chair, and sofa | intersecting faces: 56.1 % -> 0 %; 58.8 % -> 0 %; 77.0 % -> 0 % | Helped |
| Replace dense vertex decoder with pushing parametrization (ablation) | Textured 3D mesh model · ShapeNet airplane | IS: IS: 3.7; FID: 76.3; KID: 0.067 | Hurt |
| Replace dense vertex decoder with pushing parametrization (ablation) | Textured 3D mesh model · ShapeNet chair | IS: IS: 5.4; FID: 145.8; KID: 0.127 | Hurt |
| Replace dense vertex decoder with pushing parametrization (ablation) | Textured 3D mesh model · ShapeNet sofa | IS: IS: 4.0; FID: 82.6; KID: 0.058 | Hurt |
| Use a 3D textured-mesh generative model instead of a 2D image VAE | Natural-image bird generation · Birds, no-mask | FID: 213.8 -> 75.4 | Helped |
| Use a 3D textured-mesh generative model instead of a 2D image VAE | Natural-image car generation · Cars, no-mask | FID: 157.4 -> 182.1 | Hurt |
Closest papers by meaning
- Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors (ICLR 2024 · similarity 0.90)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.89)
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (NeurIPS 2025 · similarity 0.88)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 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 0.88)
- SCULPT: Shape-Conditioned Unpaired Learning of Pose-dependent Clothed and Textured Human Meshes (CVPR 2024 · similarity 0.88)
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.