Research map / Label-efficient and robust learning / Noisy labels
Seeing Through the Human Reporting Bias: Visual Classifiers From Noisy Human-Centric Labels
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 |
|---|---|---|---|
| Decouple visual presence from human-centric relevance | MILVC with VGG16 · MS COCO captions test (20k images; 1k concepts) | mAP: 34.0 -> 36.3 | Helped |
| Add image-conditioned relevance modeling (ablation) | Latent model · MS COCO 1k visual concepts | mAP: 34.2 -> 35.1 | Helped |
| Use human-centric output h instead of visual-presence output v | Latent model · MS COCO detection test (73 classes) | mAP: 66.8 -> 66.5 | Hurt |
| Add latent factorization rather than directly predicting noisy tags | MILVC · YFCC100M test (1k tags) | mAP: 5.7 -> 11.2 | Helped |
| Use latent model concept probabilities as captioner input | LSTM captioner · MS COCO caption generation | BLEU-4: 27.7 -> 29.2 | Helped |
| Estimate relevance without conditioning on the image (ablation) | Latent model · MS COCO 1k visual concepts | mAP: 35.1 -> 34.2 | Hurt |
| Replace latent model with extra fc8 layers at matched parameter count (ablation) | MILVC and Classif. baselines · MS COCO captions test (20k images; 1k concepts) | mAP: 34.0 -> 33.8 | No clear effect |
Closest papers by meaning
- Classifier-to-Bias: Toward Unsupervised Automatic Bias Detection for Visual Classifiers (CVPR 2025 · similarity 0.88)
- Modeling Collaborator: Enabling Subjective Vision Classification With Minimal Human Effort via LLM Tool-Use (CVPR 2024 · similarity 0.86)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.86)
- Discovering and Mitigating Visual Biases through Keyword Explanation (CVPR 2024 · similarity 0.86)
- Doomed to Re-Annotate, Forever: The ImageNet Story (NeurIPS 2026 · similarity 0.86)
- Common Sense Bias Modeling for Classification Tasks (AAAI 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.