Research map / Label-efficient and robust learning / Noisy labels
Noisy Label Recovery for Shadow Detection in Unfamiliar Domains
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train on lazily annotated images from the target domain instead of precisely annotated images from other domains | Shadow detection with a Least Squares SVM · SBU test | AP: 55.1 and 68.8 -> 77.5 | Helped |
| Recover labels in the SBU lazy training set | SBU-Lazy · SBU test | AP: 77.5 -> 80.1 | Helped |
| Add label recovery to UIUC lazy labels (ablation) | UIUC shadow classifier · UIUC test | AP: 79.5 -> 87.2 | Helped |
| Add label recovery to UCF lazy labels (ablation) | UCF shadow classifier · UCF test | AP: 73.5 -> 75.6 | Helped |
| Use lazy labels instead of carefully annotated labels | UIUC shadow classifier · UIUC test | AP: 88.5 vs 79.5 | Hurt |
| Use lazy labels instead of carefully annotated labels | UCF shadow classifier · UCF test | AP: 74.5 vs 73.5 | Hurt |
| Use the proposed label recovery instead of noise-tolerant C-SVM | Shadow detection with noisy UIUC labels · UIUC test | AP: 81.5 vs 87.2 | Helped |
| Use the proposed label recovery instead of noise-tolerant C-SVM | Shadow detection with noisy UCF labels · UCF test | AP: 74.3 vs 75.6 | Helped |
| Apply label recovery with positive-label noise up to the reported threshold (ablation) | Shadow classifier trained on polluted labels · UIUC and UCF training data | AP: 30% | Helped |
| Apply label recovery with negative-label noise up to the reported threshold (ablation) | Shadow classifier trained on polluted labels · UIUC and UCF training data | AP: 50% | Helped |
Closest papers by meaning
- A Multi-Task Mean Teacher for Semi-Supervised Shadow Detection (CVPR 2020 · similarity 0.86)
- S²Teacher: Step-by-step Teacher for Sparsely Annotated Oriented Object Detection (AAAI 2026 · similarity 0.85)
- ELDET: Early-Learning Distillation with Noisy Labels for Object Detection (NeurIPS 2025 · similarity 0.84)
- Video Shadow Detection via Spatio-Temporal Interpolation Consistency Training (CVPR 2022 · similarity 0.84)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.84)
- Doomed to Re-Annotate, Forever: The ImageNet Story (NeurIPS 2026 · similarity 0.83)
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.