Research map / Detection and segmentation / CNN features
The Treasure Beneath Multiple Annotations: An Uncertainty-Aware Edge Detector
Read the paperCheck 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 |
|---|---|---|---|
| Model label ambiguity as a learned distribution with an uncertainty branch (ablation) | Edge detector with baseline backbone · BSDS500 SS | ODS: 0.824 -> 0.825 | Helped |
| Supervise predicted uncertainty using variance across annotations (ablation) | UAED · BSDS500 SS | ODS: 0.825 -> 0.829 | Helped |
| Progressively weight high-uncertainty pixels more during training (ablation) | UAED with uncertainty and variance supervision · BSDS500 SS | ODS: 0.829 -> 0.829 | No clear effect |
| Use traditional weighting that down-weights high-uncertainty pixels (ablation) | UAED · BSDS500 SS | ODS: 0.825 | Hurt |
| Use fixed rather than progressive uncertainty weighting (ablation) | UAED · BSDS500 SS | ODS: 0.826 vs 0.829 | Hurt |
| Use label variance rather than predicted uncertainty for weighting (ablation) | UAED · BSDS500 SS | ODS: 0.826 vs 0.829 | Hurt |
| Use CVAE-based uncertainty estimation instead of label-space uncertainty (ablation) | Baseline edge detector · BSDS500 SS | ODS: 0.824 -> 0.821 | Hurt |
| Use MC Dropout uncertainty estimation (ablation) | Baseline edge detector · BSDS500 SS | ODS: 0.824 -> 0.825 | Helped |
| Apply UAED to VGG / UNet++ (ablation) | VGG / UNet++ · BSDS500 SS | ODS: 0.801 -> 0.808 | Helped |
| Apply UAED to EfficientNet / UNet (ablation) | EfficientNet / UNet · BSDS500 SS | ODS: 0.821 -> 0.824 | Helped |
| Apply UAED to SegFormer / UNet (ablation) | SegFormer / UNet · BSDS500 SS | ODS: 0.822 -> 0.828 | Helped |
| Use UAED rather than EDTER | Edge detection · Multicue edge | ODS: .894 (0.005) -> .895 (0.002) | Helped |
Closest papers by meaning
- SAUGE: Taming SAM for Uncertainty-Aligned Multi-Granularity Edge Detection (AAAI 2025 · similarity 0.89)
- RankED: Addressing Imbalance and Uncertainty in Edge Detection Using Ranking-based Losses (CVPR 2024 · similarity 0.89)
- DiffusionEdge: Diffusion Probabilistic Model for Crisp Edge Detection (AAAI 2024 · similarity 0.84)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.84)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.83)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · 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.