Research map / Learning theory and algorithms / Prediction and decision losses
Enhancing Image-Conditional Coverage in Segmentation: Adaptive Thresholding via Differentiable Miscoverage Loss
Rui Luo, Jie Bao, Xiaoyi Su, Wen Jung Li, Suqun Cao
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 |
|---|---|---|---|
| replace global CRC threshold with COAT image-adaptive threshold | CRC · Polyp PSPNet alpha=0.1 | coverage gap: 0.150 -> 0.110 | Improved |
| replace global CRC threshold with COAT image-adaptive threshold | CRC · Polyp PSPNet alpha=0.2 | coverage gap: 0.249 -> 0.144 | Improved |
| replace global CRC threshold with COAT image-adaptive threshold | CRC · Skin PSPNet alpha=0.2 | coverage gap: 0.103 -> 0.064 | Improved |
| replace AA-CRC supervised thresholding with COAT end-to-end loss | AA-CRC · Polyp PSPNet alpha=0.2 | coverage gap: 0.162 -> 0.144 | Improved |
| replace supervised regression (AT) with differentiable miscoverage loss (COAT) (ablation) | AT · Skin PSPNet alpha=0.1 | coverage gap: 0.061 -> 0.050 | Improved |
| replace AT with COAT (ablation) | AT · Fire PSPNet alpha=0.1 | coverage gap: 0.065 -> 0.060 | Improved |
| replace COAT with CRC (adaptive thresholds vs global) | COAT · Fire PSPNet alpha=0.1 marginal coverage | marginal coverage: 0.901 -> 0.900 | No clear effect |
| lower softmax temperature from T=100 to T=0.05 (ablation) | COAT · Polyp alpha=0.1 | coverage gap: 0.151 -> 0.102 | Improved |
| lower temperature from T=0.05 to T=0.001 (ablation) | COAT · Polyp alpha=0.1 | coverage gap: 0.102 -> 0.147 | Hurt |
| lower temperature from T=100 to T=0.05 (ablation) | COAT · Skin alpha=0.1 | coverage gap: 0.081 -> 0.055 | Improved |
| use COAT instead of AT for training speed | AT · Polyp training time PSPNet | training time (s): 151.7819 -> 24.3318 | Improved |
| use COAT instead of AA-CRC for calibration | AA-CRC · Polyp calibration time PSPNet | calibration time (s): 120.8917 -> 0.3764 | Improved |
Closest papers by meaning
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.85)
- CONSIGN: Conformal Segmentation Informed by Spatial Groupings via Decomposition (ICLR 2026 · similarity 0.85)
- Adaptive Sensitivity Analysis for Robust Augmentation against Natural Corruptions in Image Segmentation (ICML 2025 · similarity 0.84)
- Uncertainty Estimation for Safety-critical Scene Segmentation via Fine-grained Reward Maximization (NeurIPS 2023 · similarity 0.84)
- Test-time Adaptation for Foundation Medical Segmentation Model Without Parametric Updates (ICCV 2025 · similarity 0.84)
- On the Relaxation of Conditional Independence Assumption for Image Segmentation (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.
