Research map / 3D vision / Face and shape
Relightable Neural Human Assets From Multi-View Gradient Illuminations
Taotao Zhou, Kai He, Di Wu, Teng Xu, Qixuan Zhang, Kuixiang Shao, Wenzheng Chen, Lan Xu, Jingyi Yu
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 |
|---|---|---|---|
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | PSNR: 30.529 -> 30.868 | Improved |
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | SSIM: 0.968 -> 0.972 | Improved |
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | RMSE: 0.030 -> 0.029 | Improved |
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | Mean angular error: 4.314° -> 4.296° | Improved |
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | Normal error <5°: 79.891% -> 80.848% | Improved |
| Use depth-guided reprojection of captured normal and albedo maps instead of Instant-NGP volume rendering (ablation) | Neural human asset G-buffer synthesis · Held-out captured views | Normal error <25°: 96.013% -> 96.293% | Improved |
| Train relighting networks with UltraStage-rendered data rather than prior training data | Single-image human relighting · Human relighting test images | User preference over RH: 96.6 % | Improved |
| Train relighting networks with UltraStage-rendered data rather than prior training data | Single-image human relighting · Human relighting test images | User preference over RW: 83.3 % | Improved |
| Use photometric-stereo normal maps to guide geometry reconstruction instead of RGB supervision (ablation) | Neural human geometry reconstruction · Captured human scenes | Fine-detail reconstruction | Improved |
Closest papers by meaning
- Comprehensive Relighting: Generalizable and Consistent Monocular Human Relighting and Harmonization (CVPR 2025 · similarity 0.89)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · similarity 0.88)
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (CVPR 2025 · similarity 0.88)
- ReCap: Better Gaussian Relighting with Cross-Environment Captures (CVPR 2025 · similarity 0.88)
- ROGR: Relightable 3D Objects using Generative Relighting (NeurIPS 2025 · similarity 0.87)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 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.
