Research map / 3D vision / Face and shape
DiffTrans: Differentiable Geometry-Materials Decomposition for Reconstructing Transparent Objects
Changpu Li, Shuang Wu, Songlin Tang, Guangming Lu, Jun Yu, Wenjie Pei
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 tone regularization L_tone (ablation) | DiffTrans · NVS hand | PSNR: 22.37 -> 23.39 | Improved |
| add tone regularization L_tone (ablation) | DiffTrans · NVS (5 scenes, PSNR/SSIM/LPIPS) | PSNR, SSIM, LPIPS: monkey PSNR 24.15 -> 24.62; SSIM drops 0.8942 -> 0.8935 on monkey only | Improved |
| add tone regularization L_tone (ablation) | DiffTrans · NVS horse and mouse | PSNR: 27.00 -> 27.03 (horse); 24.15 -> 24.16 (mouse) | No clear effect |
| refine geometry with recursive differentiable ray tracer (stage 2) after mask-based initialization (ablation) | DiffTrans · Synthetic geometry (6 scenes) | Chamfer distance (x10^-4): 4.666 -> 3.264 | Improved |
| refine geometry with ray tracer (stage 2) (ablation) | DiffTrans · Synthetic geometry (6 scenes) | F1 (x10^-1): 8.088 -> 8.386 | Improved |
| model refraction, absorption and geometry jointly (DiffTrans vs NeRRF) | relighting · Relighting (synthetic) | PSNR: 19.25 -> 23.17 | Improved |
| DiffTrans vs NeRO for relighting SSIM | relighting · Relighting (synthetic) | SSIM: 0.8500 -> 0.8380 | Hurt |
Closest papers by meaning
- Differentiable Neural Surface Refinement for Modeling Transparent Objects (CVPR 2024 · similarity 0.89)
- Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (CVPR 2024 · similarity 0.88)
- Subsurface Scattering for Gaussian Splatting (NeurIPS 2024 · similarity 0.88)
- From Transparent to Opaque: Rethinking Neural Implicit Surfaces with $\alpha$-NeuS (NeurIPS 2024 · similarity 0.87)
- Factored-NeuS: Reconstructing Surfaces, Illumination, and Materials of Possibly Glossy Objects (CVPR 2025 · similarity 0.87)
- Multi-times Monte Carlo Rendering for Inter-reflection Reconstruction (NeurIPS 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.
