Research map / 3D vision / 3D generation
MatMart: Material Reconstruction of 3D Objects via Diffusion
Xiuchao Wu, Pengfei Zhu, Jiangjing Lyu, Xinguo Liu, Jie Guo, Yanwen Guo, Weiwei Xu, Chengfei Lyu
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 |
|---|---|---|---|
| add view-material cross-attention (VMCA) to progressive inference (ablation) | MatMart · Objaverse 3-view, 1024x1024 | Rendering FID: 28.58 -> 26.20 | Improved |
| add material priors (baked UV materials projected to view) to generation (ablation) | MatMart · Objaverse 3-view, 1024x1024 | Rendering FID: 34.93 -> 26.20 | Improved |
| add stage-1 prediction and texture baking before generation (ablation) | MatMart · Objaverse 3-view, 1024x1024 | Rendering FID: 42.40 -> 26.20 | Improved |
| increase inference resolution from 512 to 1024 (ablation) | MatMart · Objaverse 3-view | Rendering FID: 31.34 -> 26.20 | Improved |
| increase inference resolution from 512 to 1024 (ablation) | MatMart · Objaverse single-view | Rendering FID: 31.64 -> 31.49 | No clear effect |
| increase number of input views (ablation) | MatMart · Objaverse, 1024x1024 | FID: 37.39 -> 29.05 -> 27.95 -> 27.71 (1, 5, 10, 20 views) | Improved |
| increase number of input views (ablation) | MatMart · Objaverse, 1024x1024 | peak GPU memory (GB): 24.03 for all view counts | No clear effect |
| use stage-1 prediction + UV-space TexGEN generation instead of view-space generation | MatMart · Objaverse single-view | Rendering FID: 31.64 -> 54.10 (vs MatMart 512) | Hurt |
| replace Material Anything's RGB-then-predict pipeline with predict-then-generate materials | MatMart · Objaverse single-view | Rendering FID: 54.16 -> 31.64 | Improved |
| change generation group size (ablation) | MatMart · Objaverse | reconstruction quality | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MaterialMVP — MaterialMVP: Illumination-Invariant Material Generation via Multi-view PBR Diffusion (ICCV 2025) · better on 10 of 10 shared results
- Material Anything — Material Anything: Generating Materials for Any 3D Object via Diffusion (CVPR 2025) · better on 2 of 2 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
- Material Anything: Generating Materials for Any 3D Object via Diffusion (CVPR 2025 · similarity 0.88)
- Intrinsic Image Diffusion for Indoor Single-view Material Estimation (CVPR 2024 · similarity 0.87)
- Material Palette: Extraction of Materials from a Single Image (CVPR 2024 · similarity 0.86)
- IDArb: Intrinsic Decomposition for Arbitrary Number of Input Views and Illuminations (ICLR 2025 · similarity 0.86)
- Cycle3D: High-quality and Consistent Image-to-3D Generation via Generation-Reconstruction Cycle (AAAI 2025 · similarity 0.86)
- MAGE : Single Image to Material-Aware 3D via the Multi-View G-Buffer Estimation Model (CVPR 2025 · similarity 0.86)
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.
