Local Temperature Scaling for Probability Calibration

TL;DR

This paper introduces a learning-based local temperature scaling method for probability calibration in semantic segmentation, improving the reliability of predicted probabilities without affecting accuracy across multiple datasets.

Contribution

It proposes a convolutional neural network approach for local probability calibration in semantic segmentation, a novel application beyond classification tasks.

Findings

Improved calibration performance on COCO, CamVid, and LPBA40 datasets.

Effective calibration in multi-atlas brain segmentation from MRI.

Calibration as a post-processing step without accuracy loss.

Abstract

For semantic segmentation, label probabilities are often uncalibrated as they are typically only the by-product of a segmentation task. Intersection over Union (IoU) and Dice score are often used as criteria for segmentation success, while metrics related to label probabilities are not often explored. However, probability calibration approaches have been studied, which match probability outputs with experimentally observed errors. These approaches mainly focus on classification tasks, but not on semantic segmentation. Thus, we propose a learning-based calibration method that focuses on multi-label semantic segmentation. Specifically, we adopt a convolutional neural network to predict local temperature values for probability calibration. One advantage of our approach is that it does not change prediction accuracy, hence allowing for calibration as a post-processing step. Experiments on the COCO, CamVid, and LPBA40 datasets demonstrate improved calibration performance for a range of different metrics. We also demonstrate the good performance of our method for multi-atlas brain segmentation from magnetic resonance images.