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

2024-12-09

arXiv.org

AI Summary

Multi-class semantic segmentation suffers from high annotation costs and inaccurate boundary predictions, while existing patch-based active learning methods often neglect uncertainty modeling for boundary pixels. This paper proposes OREAL, a boundary-aware and class-balanced patch-level active learning framework. Its core contributions are: (1) a max-aggregated pixel-wise uncertainty metric that explicitly enhances sensitivity to object boundaries; and (2) One-vs-Rest entropy, which decouples inter-class uncertainty and implicitly enforces class-balanced sample selection. Extensive experiments across multiple benchmarks and state-of-the-art segmentation architectures demonstrate that OREAL significantly improves annotation efficiency and overall segmentation accuracy—particularly yielding substantial gains in boundary prediction quality.

Technology Category

• Machine Learning: Active Learning
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

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.

Problem

• Addresses multi-class semantic segmentation challenges in computer vision.
• Reduces dataset creation effort using Active Learning with strategic annotation.
• Improves boundary detection and class balancing in segmentation tasks.

Innovation

• OREAL enhances boundary detection via pixel-wise uncertainty aggregation.
• Introduces one-vs-rest entropy for class-wise uncertainty computation.
• Achieves implicit class balancing during dataset creation.