Active Learning with Context Sampling and One-vs-Rest Entropy for Semantic Segmentation

TL;DR

This paper introduces OREAL, a novel active learning method for semantic segmentation that improves boundary detection and class balancing through a new uncertainty measure, reducing annotation effort.

Contribution

OREAL is the first patch-based active learning approach that explicitly enhances boundary detection and incorporates one-vs-rest entropy for better class-wise uncertainty estimation.

Findings

OREAL outperforms existing methods in boundary detection accuracy.

The one-vs-rest entropy improves class balance during dataset creation.

Experiments validate OREAL's effectiveness across multiple datasets and models.

Abstract

Multi-class semantic segmentation remains a cornerstone challenge in computer vision. Yet, dataset creation remains excessively demanding in time and effort, especially for specialized domains. Active Learning (AL) mitigates this challenge by selecting data points for annotation strategically. However, existing patch-based AL methods often overlook boundary pixels critical information, essential for accurate segmentation. We present OREAL, a novel patch-based AL method designed for multi-class semantic segmentation. OREAL enhances boundary detection by employing maximum aggregation of pixel-wise uncertainty scores. Additionally, we introduce one-vs-rest entropy, a novel uncertainty score function that computes class-wise uncertainties while achieving implicit class balancing during dataset creation. Comprehensive experiments across diverse datasets and model architectures validate our hypothesis.