[CVPR 2026 Highlight] MatLat: Material Latent Space for PBR Texture Generation

Kyeongmin Yeo,   Yunhong Min,   Jaihoon Kim,   Minhyuk Sung

KAIST

       

PBR Textures Generated by MatLat. Our method produces PBR textures that accurately represent rough materials (left), metallic surfaces (middle), and complex mixed materials (right).

---

🔥 News

• [2026.04] MatLat has been selected as a Highlight at CVPR 2026.
• [2024.12] Preprint released on arXiv.

---

🎨 Qualitative Results

For more video results and side-by-side comparisons with prior methods, visit the project page.

---

📊 Quantitative Results

Comparison with PBR texture generation methods. Best in bold, runner-up underlined. FID is computed in CLIP embedding space. SC = trained from scratch, SDS = Score Distillation Sampling, MV = multi-view diffusion.

Method	Type	Shaded FID↓	Shaded KID↓	Shaded CLIP↑	Albedo FID↓	Albedo KID↓	Albedo CLIP↑	Rough. RMSE↓	Metal. RMSE↓	Time↓
MeshGen	SC	8.637	11.105	0.282	11.322	16.193	0.282	0.143	0.201	195s
TexGaussian	SC	6.025	3.571	0.301	12.119	9.381	0.299	0.145	0.243	73s
Paint-it	SDS	8.547	5.382	0.309	13.063	10.665	0.299	0.168	0.200	1260s
DreamMat	SDS	5.422	2.668	0.311	9.621	6.002	0.311	0.167	0.165	2400s
FlashTex	SDS	7.119	5.354	0.305	12.320	10.441	0.298	0.143	0.186	285s
MaterialAnything	MV	6.582	5.287	0.312	12.691	9.325	0.317	0.233	0.200	500s
MaterialMVP	MV	6.309	5.744	0.294	9.630	8.811	0.290	0.175	0.133	35s
MatLat (Ours)	MV	3.083	1.327	0.318	4.599	1.574	0.314	0.158	0.134	34s

---

📝 Citation

@inproceedings{yeo2026matlat,
  title     = {MatLat: Material Latent Space for PBR Texture Generation},
  author    = {Yeo, Kyeongmin and Min, Yunhong and Kim, Jaihoon and Sung, Minhyuk},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2026},
}