Research map / Image generation / Autoregressive and latent generation
MatLat: Material Latent Space for PBR Texture Generation
Kyeongmin Yeo, Yunhong Min, Jaihoon Kim, Minhyuk Sung
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 |
|---|---|---|---|
| fine-tune VAE with residual prediction and KL regularization instead of frozen VAE (ablation) | MATLAT · PBR ablation | shaded FID: 3.419 -> 3.083 | Improved |
| fine-tune VAE with residual prediction and KL regularization instead of frozen VAE (ablation) | MATLAT · PBR ablation | shaded KID: 1.052 -> 1.327 | Hurt |
| single-latent MATVAE instead of separate branches (ablation) | MATLAT · PBR ablation | generation time: 111s -> 34s | Improved |
| use KL regularization instead of identity loss (ablation) | MATVAE · PBR ablation | shaded FID: 3.210 -> 3.083 | Improved |
| use residual instead of direct prediction (ablation) | MATVAE · PBR ablation | shaded FID: 3.192 -> 3.083 | Improved |
| use residual instead of direct prediction (ablation) | MATVAE · PBR ablation | metallic RMSE: 0.170 -> 0.134 | Improved |
| add locality regularization (ablation) | CAA with MATVAE · PBR ablation | c-PSNR: 19.437 -> 21.934 | Improved |
| add locality regularization (ablation) | CAA with MATVAE · PBR ablation | albedo FID: 5.873 -> 4.599 | Improved |
| add correspondence-aware attention (ablation) | MATVAE with Llocal · PBR ablation | c-PSNR: 18.687 -> 21.934 | Improved |
| add correspondence-aware attention (ablation) | MATVAE with Llocal · PBR ablation | shaded FID: 3.110 -> 3.083 | No clear effect |
| MATLAT vs SDS-based DreamMat | MATLAT · PBR texture generation | shaded FID: 5.422 -> 3.083 | Improved |
| MATLAT vs MaterialAnything on albedo CLIP | MATLAT · PBR texture generation | albedo CLIP: 0.317 -> 0.314 | Hurt |
Closest papers by meaning
- IntrinsiX: High-Quality PBR Generation using Image Priors (NeurIPS 2025 · similarity 0.89)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.89)
- MeshGen: Generating PBR Textured Mesh with Render-Enhanced Auto-Encoder and Generative Data Augmentation (CVPR 2025 · similarity 0.89)
- TexGaussian: Generating High-quality PBR Material via Octree-based 3D Gaussian Splatting (CVPR 2025 · similarity 0.88)
- PBR-SR: Mesh PBR Texture Super Resolution from 2D Image Priors (NeurIPS 2025 · similarity 0.88)
- Material Palette: Extraction of Materials from a Single Image (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.
