Research map / 3D vision / Depth and stereo
DeepShadow: Neural Shape from Shadow
Asaf Karnieli, Ohad Fried, Yacov Hel-Or
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 |
|---|---|---|---|
| recover depth from shadow maps only instead of shape-from-shading (Peng et al.) | DeepShadow · Rendered objects, average | nMZE: 0.31 -> 0.18 | Improved |
| DeepShadow instead of Santo et al. on depth | DeepShadow · Rendered objects, average | nMZE: 0.87 -> 0.18 | Improved |
| DeepShadow instead of Chen et al. on normals | DeepShadow · Rendered objects, average | MAE: 23.14 -> 22.63 | No clear effect |
| DeepShadow instead of Santo et al. on normals | DeepShadow · Rendered objects, average | MAE: 35.93 -> 22.63 | Improved |
| shadow-only DeepShadow on concave vase vs Kaya et al. | DeepShadow · Dome dataset, Vase | MAE: 16.40 -> 11.24 | Improved |
| shadow-only DeepShadow on sparse-shadow golf ball and face | DeepShadow · Dome dataset, Golf Ball | MAE: 14.23 -> 16.51 | Hurt |
| increase average shadow amount from 48% to 69% (ablation) | DeepShadow · Surface object, MAE | MAE: 20.43 -> 17.53 | Improved |
| increase average shadow amount from 48% to 69% (ablation) | Santo et al. · Surface object, MAE | MAE: 16.18 -> 18.25 | Hurt |
| add specular material to rose object (ablation) | DeepShadow · Rose, normals | MAE: 26.50 -> 27.76 | Hurt |
| add specular material to rose object (ablation) | SDPS · Rose, normals | MAE: 25.02 -> 34.35 | Hurt |
| DeepShadow instead of Chen et al. when lights at varying distances | DeepShadow · Object E (60%, varying light distance) | MAE: 22.99 -> 18.68 | Improved |
Later papers that compared with this method
1 later accepted paper reports DeepShadow as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ShadowNeuS: Neural SDF Reconstruction by Shadow Ray Supervision (CVPR 2023 · better on 4 of 4 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
- Towards Learning Neural Representations from Shadows (ECCV 2022 · similarity 0.90)
- DeepPS2: Revisiting Photometric Stereo Using Two Differently Illuminated Images (ECCV 2022 · similarity 0.89)
- Neural Reflectance for Shape Recovery With Shadow Handling (CVPR 2022 · similarity 0.89)
- Front2Back: Single View 3D Shape Reconstruction via Front to Back Prediction (CVPR 2020 · similarity 0.88)
- S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint (NeurIPS 2022 · similarity 0.87)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · similarity 0.86)
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.
