Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Segment Anything with Robust Uncertainty-Accuracy Correlation
Hongyou Zhou, Marc Toussaint, Ling Shao, Zihan Ye
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 Bayesian uncertainty estimation decoder | SAM2-FT · Zero-shot DG, 23 datasets | J&F avg: 79.75 -> 79.87 | No clear effect |
| add adversarial uncertainty estimation (style + deform) to Bayes-SAM2 (ablation) | Bayes-SAM2 · Zero-shot DG, 23 datasets | J&F avg: 79.87 -> 81.62 | Improved |
| add deformation attack to style-GCN attack (ablation) | RUAC-Style-GCN · Zero-shot DG | J&F avg: 80.74 -> 81.62 | Improved |
| add style attack to deformation-only attack (ablation) | RUAC-Deform · Zero-shot DG | J&F avg: 80.44 -> 81.62 | Improved |
| add GCN refinement to multi-object style attack (ablation) | RUAC-Style-Multi · Zero-shot DG | J&F avg: 80.52 -> 80.74 | Improved |
| add background to multi-object style perturbation (ablation) | RUAC-Style-Multi · Zero-shot DG | J&F avg: 80.52 -> 80.68 | Improved |
| perturb multiple objects instead of target only (ablation) | RUAC-Style-Single · Zero-shot DG | J&F avg: 80.43 -> 80.52 | Improved |
| replace global style perturbation with per-object style (ablation) | RUAC-Style-Global · Zero-shot DG | J&F avg: 80.26 -> 80.52 | Improved |
| add adversarial uncertainty estimation (ablation) | Bayes-SAM2 · OOD calibration | PAvPU: 55.40 -> 63.72 | Improved |
| add adversarial uncertainty estimation (ablation) | Bayes-SAM2 · In-domain MOSE | PAvPU: 86.9 -> 81.6 | Hurt |
| add adversarial uncertainty estimation (ablation) | Bayes-SAM2 · OOD calibration | ECE: 0.123 -> 0.117 | Improved |
| apply uncertainty-guided correction filter | RUAC · 23 OOD datasets | J&F: 81.62 -> 81.71 | Improved |
Closest papers by meaning
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.86)
- RobustSAM: Segment Anything Robustly on Degraded Images (CVPR 2024 · similarity 0.85)
- Benchmarking Segmentation Models with Mask-Preserved Attribute Editing (CVPR 2024 · similarity 0.83)
- BA-SAM: Scalable Bias-Mode Attention Mask for Segment Anything Model (CVPR 2024 · similarity 0.83)
- Adaptive Sensitivity Analysis for Robust Augmentation against Natural Corruptions in Image Segmentation (ICML 2025 · similarity 0.82)
- Unsegment Anything by Simulating Deformation (CVPR 2024 · similarity 0.82)
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.
