Research map / 3D vision / Depth and stereo
Cast and Attached Shadow Detection via Iterative Light and Geometry Reasoning
Shilin Hu, Jingyi Xu, Sagnik Das, Dimitris Samaras, Hieu Le
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 surface normal maps as input (ablation) | Ours · Attached shadow | BER: 20.87 -> 12.92 | Improved |
| add iterative feedback of partial attached map (ablation) | Ours · Attached shadow | BER: 15.10 -> 12.92 | Improved |
| increase iterations from 3 to 5 (ablation) | Ours · Full and attached shadow | BER: full 6.50 -> 6.11; attached 12.92 -> 13.51 | Mixed |
| use 2 iterations rather than 1 (non-iterative) (ablation) | Ours · Attached shadow | BER: 15.10 -> 13.22 | Improved |
| add margin loss L_dist (ablation) | Ours · Attached shadow | BER: 13.45 -> 12.92 | Improved |
| add L_att loss (ablation) | Ours · Attached shadow | BER: 14.61 -> 12.92 | Improved |
| add L_dir loss (ablation) | Ours · Attached shadow | BER: 15.26 -> 12.92 | Improved |
| fine-tune baselines on the new dataset with full-shadow supervision | SILT · Attached shadow | BER: 32.70 -> 26.57 | Improved |
| geometry/light-aware iterative framework vs best retrained baseline | SILT backbone · Attached shadow | BER: 19.49 -> 12.92 | Improved |
| iterative framework on cast shadow BER vs fine-tuned SILT | Ours · Cast shadow | BER: 4.52 -> 5.04 | No clear effect |
| 3 iterations vs 1 in light estimation (ablation) | Joint framework · WSRD light direction | MAE: 0.061 -> 0.049 | Improved |
| use separate attached mask to refine shadow removal | ShadowFormer · User study | preference (%): 80.7% | Improved |
Closest papers by meaning
- Unveiling Deep Shadows: A Survey and Benchmark on Image and Video Shadow Detection, Removal, and Generation in the Deep Learning Era (IJCV 2026 · similarity 0.91)
- OmniSR: Shadow Removal Under Direct and Indirect Lighting (AAAI 2025 · similarity 0.88)
- MetaShadow: Object-Centered Shadow Detection, Removal, and Synthesis (CVPR 2025 · similarity 0.88)
- Recasting Regional Lighting for Shadow Removal (AAAI 2024 · similarity 0.87)
- Shadow-Enlightened Image Outpainting (CVPR 2024 · similarity 0.87)
- Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (CVPR 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.
