Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond In-Domain Scenarios: Robust Density-Aware Calibration
Christian Tomani, Futa Kai Waseda, Yuesong Shen, Daniel Cremers
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 DAC before post-hoc calibration | Post-hoc calibration methods on CIFAR-C and ImageNet-C · CIFAR-C and ImageNet-C, severity 0–5 | ECE (×10²): 4.96 -> 4.39 | Improved |
| Add DAC before post-hoc calibration | Post-hoc calibration methods on ImageNet-C · ImageNet-C, severity 5 | ECE (×10²): -1.25 | Improved |
| Add DAC before post-hoc calibration | Post-hoc calibration methods on ImageNet-1k / ObjectNet-OOD · ImageNet-1k / ObjectNet-OOD, DenseNet169 | AUROC: 84.92 -> 85.50 | Improved |
| Use all classifier layers instead of the selected layer subset (ablation) | DAC with ResNet18 · CIFAR10-C | ECE | No clear effect |
| Vary the KNN neighborhood size k (ablation) | DAC combined with ETS or SPL · CIFAR-C and ImageNet-C | ECE (×10²) | No clear effect |
| Use the recommended KNN neighborhood size (ablation) | DAC · CIFAR-C and ImageNet-C | ECE (×10²) | Improved |
| Reduce the validation set size (ablation) | DAC combined with ETS or SPL · ImageNet DenseNet169, validation sizes 10%–100% | ECE (×10²) | No clear effect |
| Use DAC with ViT-B and ETS instead of ETS alone | ETS on ImageNet-C · ImageNet-C, ViT-B | ECE: 12% | Improved |
Closest papers by meaning
- Post-Hoc Uncertainty Calibration for Domain Drift Scenarios (CVPR 2021 · similarity 0.91)
- Uncertainty Calibration with Energy Based Instance-wise Scaling in the Wild Dataset (ECCV 2024 · similarity 0.89)
- Improving model calibration with accuracy versus uncertainty optimization (NeurIPS 2020 · similarity 0.89)
- A Benchmark Study on Calibration (ICLR 2024 · similarity 0.88)
- Improving Accuracy and Calibration via Differentiated Deep Mutual Learning (CVPR 2025 · similarity 0.86)
- Self-Calibrating Vicinal Risk Minimisation for Model Calibration (CVPR 2024 · similarity 0.86)
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.
