Research map / 3D vision / Depth and stereo
Outdoor inverse rendering from a single image using multiview self-supervision
Ye Yu, William A. P. Smith
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 VGG perceptual loss (ablation) | InverseRenderNet-style full model · Illumination estimation (global scale) | lighting MSE: 0.053 -> 0.038 | Improved |
| add VGG perceptual loss (ablation) | full model · MegaDepth reflectance | MSE: 0.0095 -> 0.0093 | Improved |
| add VGG perceptual loss (ablation) | full model · IIW | WHDR: 21.6 -> 21.1 | Improved |
| add cross-projection (albedo consistency) loss (ablation) | full model · MegaDepth reflectance | MSE: 0.0153 -> 0.0093 | Improved |
| add cross-projection loss (ablation) | full model · Illumination estimation (global scale) | lighting MSE: 0.052 -> 0.038 | Improved |
| add cross-rendering loss (ablation) | full model · MegaDepth reflectance | MSE: 0.0101 -> 0.0093 | Improved |
| add cross-rendering loss (ablation) | full model · IIW | WHDR: 21.4 -> 21.1 | Improved |
| add cross-rendering loss (ablation) | full model · Illumination estimation (global scale) | lighting MSE: 0.052 -> 0.038 | Improved |
| add cross-projection loss (ablation) | full model · IIW | WHDR: 21.6 -> 21.1 | Improved |
| self-supervised MVS training vs supervised MegaDepth normals | ours vs MegaDepth · MegaDepth test | normal mean angular error: 43.6 -> 31.2 | Improved |
| self-supervised multiview training vs InverseRenderNet | ours · MegaDepth test | normal mean angular error: 40.0 -> 31.2 | Improved |
| tune loss weights w1,w3,w4 per task instead of uniform weights (ablation) | full model · evaluation metrics | mixed | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MegaDepth — MegaDepth: Learning Single-View Depth Prediction From Internet Photos (CVPR 2018) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- InverseRenderNet: Learning Single Image Inverse Rendering (CVPR 2019 · similarity 0.97)
- Neural Multi-View Self-Calibrated Photometric Stereo without Photometric Stereo Cues (ICCV 2025 · similarity 0.90)
- MVInverse: Feed-forward Multiview Inverse Rendering in Seconds (CVPR 2026 · similarity 0.89)
- Neural Inverse Rendering of an Indoor Scene From a Single Image (ICCV 2019 · similarity 0.89)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · similarity 0.87)
- Diffusion Reflectance Map: Single-Image Stochastic Inverse Rendering of Illumination and Reflectance (CVPR 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.
