Research map / 3D vision / 3D generation
Lafite: A Generative Latent Field for 3D Native Texturing
Chia-Hao Chen, Yuan-Chen Guo, Zi-Xin Zou, Ze Yuan, Guan Luo, Xiaojuan Qi, Ding Liang, Yan-Pei Cao, Song-Hai 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 |
|---|---|---|---|
| encode colored point cloud directly instead of projected DINO features (TRELLIS-RF128) | LaFiTe VAE · VAE reconstruction 128 | PSNR: 23.07 -> 34.62 | Improved |
| encode colored point cloud directly instead of projected DINO features (TRELLIS-GS64) | LaFiTe VAE · VAE reconstruction 64 | PSNR: 21.37 -> 33.57 | Improved |
| encode colored point cloud directly instead of projected DINO features | LaFiTe VAE · VAE reconstruction 64 | LPIPS: 0.116 -> 0.053 | Improved |
| increase input points from 20,480 to 2,048,000 (ablation) | LaFiTe VAE · Reconstruction, 128 resolution | PSNR: 26.07 -> 34.30 | Improved |
| increase input points from 204,800 to 2,048,000 (ablation) | LaFiTe VAE · Reconstruction, 64 resolution | PSNR: 32.22 -> 33.49 | Improved |
| increase points from 2,048,000 to 4,096,000 (ablation) | LaFiTe VAE · Reconstruction, 128 resolution | PSNR: 34.30 -> 34.45 | Improved |
| fine-tune VAE at 128 resolution instead of using 64 model (starred) | LaFiTe VAE · VAE reconstruction | PSNR: 34.48 -> 34.62 | Improved |
| use LaFiTe instead of UniTEX | LaFiTe · Image-conditioned (unshaded) | FID: 106.45 -> 107.46 | Hurt |
| use LaFiTe instead of MaterialMVP | LaFiTe · Image-conditioned (shaded) | FD_CLIP: 48.71 -> 46.28 | Improved |
| use LaFiTe instead of MaterialMVP | LaFiTe · Image-conditioned (shaded) | FID: 101.66 -> 101.91 | Hurt |
| use LaFiTe instead of SeqTex | LaFiTe · Image-conditioned (shaded) | KD_DINO: 0.032 -> 0.027 | Improved |
| condition on geometry latent from monochrome point cloud instead of 2D projected maps | LaFiTe · Texture generation | quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- MVPaint — MVPaint: Synchronized Multi-View Diffusion for Painting Anything 3D (CVPR 2025) · better on 3 of 3 shared results
- MaterialAnything — Material Anything: Generating Materials for Any 3D Object via Diffusion (CVPR 2025) · better on 3 of 3 shared results
- MaterialMVP — MaterialMVP: Illumination-Invariant Material Generation via Multi-view PBR Diffusion (ICCV 2025) · better on 3 of 3 shared results
- Paint3D — Paint3D: Paint Anything 3D with Lighting-Less Texture Diffusion Models (CVPR 2024) · 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
- NaTex: Seamless Texture Generation as Latent Color Diffusion (CVPR 2026 · similarity 0.89)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.88)
- TriTex: Learning Texture from a Single Mesh via Triplane Semantic Features (CVPR 2025 · similarity 0.88)
- GenesisTex: Adapting Image Denoising Diffusion to Texture Space (CVPR 2024 · similarity 0.87)
- Paint3D: Paint Anything 3D with Lighting-Less Texture Diffusion Models (CVPR 2024 · similarity 0.87)
- Paint-it: Text-to-Texture Synthesis via Deep Convolutional Texture Map Optimization and Physically-Based Rendering (CVPR 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.
