Research map / Video / Video segmentation and tracking
From Shadow Segmentation to Shadow Removal
Hieu Le, Dimitris Samaras
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 bounded search space for shadow parameters w and b (ablation) | Ours · ISTD adjusted test | RMSE LAB shadow area: 47.5 -> 9.7 | Improved |
| add boundary loss L_bd (ablation) | Ours · ISTD adjusted test | RMSE LAB shadow area: 41.7 -> 9.7 | Improved |
| add matting loss L_mat-alpha (ablation) | Ours · ISTD adjusted test | RMSE LAB shadow area: 38.7 -> 9.7 | Improved |
| add smoothness loss L_sm-alpha (ablation) | Ours · ISTD adjusted test | RMSE LAB shadow area: 10.2 -> 9.7 | Improved |
| add adversarial loss L_GAN (ablation) | Ours · ISTD adjusted test | RMSE LAB shadow area: 26.9 -> 9.7 | Improved |
| add boundary loss L_bd (ablation) | Ours · ISTD adjusted test | RMSE LAB all: 9.8 -> 4.0 | Improved |
| add matting loss L_mat-alpha (ablation) | Ours · ISTD adjusted test | RMSE LAB all: 9.0 -> 4.0 | Improved |
| train on unpaired shadow patches with only shadow masks instead of unpaired shadow-free images (MaskShadow-GAN) | Ours · ISTD adjusted test | RMSE LAB shadow area: 12.4 -> 9.7 | Improved |
| use paired shadow-free supervision (SP+M-Net) instead of mask-only training | SP+M-Net · ISTD adjusted test | RMSE LAB shadow area: 7.9 -> 9.7 | Mixed |
| fine-tune on each test video for 1 epoch with detector-generated masks | Ours · Video Shadow Removal dataset | RMSE LAB on moving-shadow mask: 20.9 -> 18.0 | Improved |
| use patch-based mask-only model rather than SP+M-Net on videos | SP+M-Net · Video Shadow Removal dataset | RMSE LAB on moving-shadow mask: 22.2 -> 20.9 | Improved |
Closest papers by meaning
- Physics-Based Shadow Image Decomposition for Shadow Removal (TPAMI 2021 · similarity 0.92)
- Mask-ShadowGAN: Learning to Remove Shadows From Unpaired Data (ICCV 2019 · similarity 0.91)
- OmniSR: Shadow Removal Under Direct and Indirect Lighting (AAAI 2025 · similarity 0.87)
- When Shadow Removal Meets Intrinsic Image Decomposition: A Joint Learning Framework Using Unpaired Data (AAAI 2025 · similarity 0.87)
- Polarization Guided Mask-Free Shadow Removal (AAAI 2025 · similarity 0.87)
- SoftShadow: Leveraging Soft Masks for Penumbra-Aware Shadow Removal (CVPR 2025 · 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.
