Research map / 3D vision / Face and shape
Material Editing Using a Physically Based Rendering Network
Guilin Liu, Duygu Ceylan, Ersin Yumer, Jimei Yang, Jyh-Ming Lien
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 rendering layer with perceptual loss to material prediction module (ablation) | material prediction module · Material prediction with GT normals and illumination | rendered image L2: 544.1 -> 355.5 | Improved |
| add rendering layer with perceptual loss to material prediction module (ablation) | material prediction module · Material prediction with GT normals and illumination | rendered image SSIM: 0.9297 -> 0.9517 | Improved |
| add rendering layer with perceptual loss to material prediction module (ablation) | material prediction module · Material prediction with GT normals and illumination | material coefficient L2: 4.1038 -> 4.8144 | Hurt |
| add rendering layer with perceptual loss to normal prediction module (ablation) | normal prediction module · Surface normal prediction (car) | rendered image L2: 313.2058 -> 238.4619 | Improved |
| add rendering layer with perceptual loss to normal prediction module (ablation) | normal prediction module · Surface normal prediction (car) | rendered image SSIM: 0.9441 -> 0.9625 | Improved |
| add rendering layer with perceptual loss to normal prediction module (ablation) | normal prediction module · Surface normal prediction (car) | normals cosine: 0.9105 -> 0.9050 | No clear effect |
| add post-optimization of intrinsic properties (ablation) | our network · Material transfer, synthetic test | L2 error: 805.2 -> 530.9 | Improved |
| add post-optimization of intrinsic properties (ablation) | our network · Material transfer, synthetic test | SSIM: 0.9408 -> 0.9557 | Improved |
| add post-optimization of intrinsic properties (ablation) | Baseline 2 (independent predictions) · Material transfer, synthetic test | L2 error: 886.9 -> 1273.7 | Hurt |
| replace independent L2-trained normal, material, illumination networks with joint network with rendering layer | material transfer pipeline · Material transfer, synthetic test | L2 error: 886.9 -> 805.2 | Improved |
| use rendering-layer network instead of encoder-decoder feature swapping | material transfer pipeline · Material transfer, synthetic test | L2 error: 1076.7 -> 805.2 | Improved |
Closest papers by meaning
- Material Palette: Extraction of Materials from a Single Image (CVPR 2024 · similarity 0.88)
- PhyS-EdiT: Physics-aware Semantic Image Editing with Text Description (CVPR 2025 · similarity 0.88)
- Alchemist: Parametric Control of Material Properties with Diffusion Models (CVPR 2024 · similarity 0.87)
- Material Anything: Generating Materials for Any 3D Object via Diffusion (CVPR 2025 · similarity 0.87)
- MatFuse: Controllable Material Generation with Diffusion Models (CVPR 2024 · similarity 0.87)
- MARBLE: Material Recomposition and Blending in CLIP-Space (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.
