Research map / 3D vision / Depth and stereo
Fine-Scale Surface Normal Estimation Using a Single NIR Image
Youngjin Yoon, Gyeongmin Choe, Namil Kim, Joon-Young Lee, In So Kweon
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 angular loss to L2 (ablation) | GAN with L2 loss · NIR dataset (all views, 3rd lighting direction) | mean angular error (deg): 16.72 -> 15.56 | Improved |
| add angular loss to L1 (ablation) | GAN with L1 loss · NIR dataset (all views, 3rd lighting direction) | mean angular error (deg): 16.82 -> 15.88 | Improved |
| add angular loss to L2 (ablation) | GAN with L2 loss · NIR dataset (center view, 3rd lighting direction) | mean angular error (deg): 8.76 -> 7.70 | Improved |
| add angular loss to L1 (ablation) | GAN with L1 loss · NIR dataset (center view, 3rd lighting direction) | mean angular error (deg): 10.01 -> 7.35 | Improved |
| add integrability loss to L2+Lang (ablation) | GAN with L2+Lang · NIR dataset (all views, 3rd lighting direction) | mean angular error (deg): 15.56 -> 15.71 | Hurt |
| add integrability loss to L1+Lang (ablation) | GAN with L1+Lang · NIR dataset (all views, 3rd lighting direction) | mean angular error (deg): 15.88 -> 16.46 | Hurt |
| add integrability loss to L2+Lang (ablation) | GAN with L2+Lang · NIR dataset (center view, 3rd lighting direction) | mean angular error (deg): 7.70 -> 4.75 | Mixed |
| add integrability loss to L1+Lang (ablation) | GAN with L1+Lang · NIR dataset (center view, 3rd lighting direction) | mean angular error (deg): 7.35 -> 9.94 | Hurt |
| use L2 instead of L1 intensity loss (ablation) | GAN · NIR dataset (center view, 3rd lighting direction) | mean angular error (deg): 10.01 -> 8.76 | Improved |
| add angular loss to L2 (ablation) | GAN with L2 loss · NIR dataset (all views, 3rd lighting direction) | good pixels within 15 deg (%): 36.12 -> 49.77 | Improved |
| evaluate detail map instead of full normal map (ablation) | L2+Lang · NIR dataset (all views, 3rd lighting direction) | mean angular error (deg): 15.56 -> 3.61 | Improved |
Closest papers by meaning
- Rethinking Inductive Biases for Surface Normal Estimation (CVPR 2024 · similarity 0.89)
- VPLNet: Deep Single View Normal Estimation With Vanishing Points and Lines (CVPR 2020 · similarity 0.88)
- Deep Iterative Surface Normal Estimation (CVPR 2020 · similarity 0.87)
- Estimating and Exploiting the Aleatoric Uncertainty in Surface Normal Estimation (ICCV 2021 · similarity 0.87)
- Normal-NeRF: Ambiguity-Robust Normal Estimation for Highly Reflective Scenes (AAAI 2025 · similarity 0.86)
- PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction (CVPR 2025 · 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.
