Research map / Detection and segmentation / CNN features
Local Temperature Scaling for Probability Calibration
Zhipeng Ding, Xu Han, Peirong Liu, Marc Niethammer
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 |
|---|---|---|---|
| use spatially varying local temperature scaling instead of a global temperature | TS · COCO FCN, All region | ECE (%): 12.53 -> 10.04 | Improved |
| use local temperature scaling instead of image-based temperature scaling | IBTS · COCO FCN, All region | ECE (%): 11.92 -> 10.04 | Improved |
| apply global temperature scaling | Uncalibrated FCN · COCO FCN, All region | ECE (%): 12.44 -> 12.53 | No clear effect |
| use LTS instead of TS | Tiramisu · CamVid, All region | ECE (%): 3.45 -> 3.40 | Improved |
| use LTS instead of TS | Tiramisu · CamVid, Boundary region | ECE (%): 12.66 -> 11.80 | Improved |
| use LTS instead of TS | U-Net · LPBA40 U-Net, All region | ECE (%): 1.43 -> 0.90 | Improved |
| use LTS instead of TS | VoteNet+ · LPBA40 VoteNet+, All region | ECE (%): 5.07 -> 0.71 | Improved |
| add LTS post-hoc calibration to MMCE training | MMCE · CamVid, All region | ECE (%): 4.45 -> 4.15 | Improved |
| add LTS post-hoc calibration to focal loss training | FL · CamVid, All region | ECE (%): 3.47 -> 3.13 | Improved |
| add LTS post-hoc calibration to focal loss training | FL · CamVid, Boundary region | ECE (%): 8.68 -> 11.06 | Hurt |
| use LTS calibrated probabilities in joint label fusion instead of uncalibrated | VoteNet+ with JLF · LPBA40 MAS | Surface Dice SD (%): 75.89 -> 75.96 | Improved |
| use LTS calibrated probabilities in joint label fusion instead of uncalibrated | VoteNet+ with JLF · LPBA40 MAS | ASD (mm): 0.99 -> 0.98 | Improved |
Closest papers by meaning
- ValUES: A Framework for Systematic Validation of Uncertainty Estimation in Semantic Segmentation (ICLR 2024 · similarity 0.84)
- On the Limitations of Temperature Scaling for Distributions with Overlaps (ICLR 2024 · similarity 0.84)
- SFC: Shared Feature Calibration in Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.84)
- On Temperature Scaling and Conformal Prediction of Deep Classifiers (ICML 2025 · similarity 0.83)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.83)
- Beyond One-Hot Labels: Semantic Mixing for Model Calibration (ICML 2025 · 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.
