Research map / 3D vision / Face and shape
Multi-View Mesh Reconstruction With Neural Deferred Shading
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use explicit triangle meshes with differentiable rasterization and neural shading instead of an implicit neural surface representation | Neural Deferred Shading · DTU MVS | optimization runtime: 80 times faster than IDR | Helped |
| Use explicit triangle meshes with differentiable rasterization and neural shading instead of an implicit neural surface representation | Neural Deferred Shading · DTU MVS | Chamfer-L1 reconstruction accuracy | No clear effect |
| Add Laplacian regularization (ablation) | Neural Deferred Shading · DTU MVS reconstruction | mesh artifacts | Helped |
| Add normal consistency regularization (ablation) | Neural Deferred Shading · DTU MVS reconstruction | smoothness | Helped |
| Add the silhouette loss (ablation) | Neural Deferred Shading · DTU MVS reconstruction | boundary reconstruction | Helped |
| Add the shading loss (ablation) | Neural Deferred Shading · DTU MVS reconstruction | geometric detail | Helped |
| Use positional encoding with 4 octaves (ablation) | Neural Deferred Shading · DTU MVS reconstruction | reconstruction accuracy: 4 octaves | Helped |
| Replace positional encoding with SIREN activations (ablation) | Neural Deferred Shading · DTU MVS reconstruction | geometric detail | Hurt |
| Replace positional encoding with Gaussian Fourier features (ablation) | Neural Deferred Shading · DTU MVS reconstruction | geometric detail | Hurt |
| Use a ReLU MLP without positional encoding (ablation) | Neural Deferred Shading · DTU MVS reconstruction | geometric detail | Mixed |
| Use fewer than 2 layers or more than 512 units per layer (ablation) | Neural Deferred Shading · DTU MVS reconstruction | geometric detail: less than 2 layers | Hurt |
| Use 3 layers with 256 units per layer (ablation) | Neural Deferred Shading · DTU MVS reconstruction | reconstruction quality: 3 layers with 256 units per layer | Helped |
Closest papers by meaning
- Differentiable Neural Surface Refinement for Modeling Transparent Objects (CVPR 2024 · similarity 0.89)
- GTR: Improving Large 3D Reconstruction Models through Geometry and Texture Refinement (ICLR 2025 · similarity 0.88)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · similarity 0.88)
- QUADify: Extracting Meshes with Pixel-level Details and Materials from Images (CVPR 2024 · similarity 0.87)
- AniSDF: Fused-Granularity Neural Surfaces with Anisotropic Encoding for High-Fidelity 3D Reconstruction (ICLR 2025 · similarity 0.87)
- GURecon: Learning Detailed 3D Geometric Uncertainties for Neural Surface Reconstruction (AAAI 2025 · 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.