Research map / 3D vision / Face and shape
Differentiable Shadow Mapping for Efficient Inverse Graphics
Markus Worchel, Marc Alexa
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 differentiable shadow mapping to rasterization | Rasterizer pose estimation · Monocular pose estimation, Dragon, 512 × 512 | rotation error (°): 8.02 -> 3.32 | Improved |
| Add differentiable shadow mapping to rasterization | Rasterizer pose estimation · Monocular pose estimation, Spot, 512 × 512 | rotation error (°): 5.80 -> 0.05 | Improved |
| Use rasterization with shadow maps instead of path tracing | Monocular pose estimation · Monocular pose estimation, Spot, 512 × 512 | runtime [s]: 381.80 -> 3.76 | Improved |
| Remove shadows and rely only on surface shading (ablation) | Differentiable shadow mapping pose estimation · Monocular pose estimation | pose accuracy | Hurt |
| Disable antialias smoothing on the shadow map and squared depth map (ablation) | Differentiable shadow mapping · Single-occluder shadow reconstruction | convergence to target configuration | Hurt |
| Apply antialias smoothing to the shadow map and squared depth map (ablation) | Differentiable shadow mapping · Single-occluder shadow reconstruction | convergence to target configuration | Improved |
| Increase the filter kernel size (ablation) | Differentiable shadow mapping · Single-occluder optimization | gradient-descent robustness | Improved |
| Increase the filter kernel size (ablation) | Differentiable shadow mapping · Light direction estimation | alignment accuracy | No clear effect |
| Increase the filter kernel size (ablation) | Differentiable shadow mapping · Light direction estimation | runtime | Hurt |
| Increase shadow map resolution (ablation) | Differentiable shadow mapping · Light direction estimation | runtime | Hurt |
| Use three shadows instead of one (ablation) | Differentiable shadow mapping face reconstruction · Synthetic face reconstruction, mean across faces | D_mean: 0.013 -> 0.008 | Improved |
| Use three shadows instead of one (ablation) | Differentiable shadow mapping face reconstruction · Synthetic face reconstruction, mean across faces | D_Hausdorff: 0.105 -> 0.068 | Improved |
Closest papers by meaning
- Differentiable Point-based Inverse Rendering (CVPR 2024 · similarity 0.88)
- Inverse Rendering using Multi-Bounce Path Tracing and Reservoir Sampling (ICLR 2025 · similarity 0.88)
- Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (CVPR 2024 · similarity 0.86)
- LinPrim: Linear Primitives for Differentiable Volumetric Rendering (NeurIPS 2025 · similarity 0.86)
- Multi-times Monte Carlo Rendering for Inter-reflection Reconstruction (NeurIPS 2024 · similarity 0.85)
- RobIR: Robust Inverse Rendering for High-Illumination Scenes (NeurIPS 2024 · similarity 0.85)
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.
