Semi-Supervised Semantic Segmentation Based on Pseudo-Labels: A Survey

2024-03-04

arXiv.org

AI Summary

This paper addresses the high cost of pixel-level annotations in semantic segmentation by systematically reviewing pseudo-labeling–driven semi-supervised learning methods. We propose the first multi-dimensional classification framework covering three sequential stages—pseudo-label generation, refinement, and utilization—to clarify technical evolution. Our synthesis integrates key techniques including consistency regularization, uncertainty modeling, curriculum learning, contrastive learning, and cross-domain adaptation, significantly enhancing pseudo-label quality and robustness. Furthermore, we construct a comprehensive knowledge graph encompassing mainstream methodologies, representative application domains (e.g., medical imaging and remote sensing), standard benchmarks (e.g., Cityscapes, PASCAL VOC), and widely adopted evaluation metrics (e.g., mIoU). The work provides theoretical foundations for algorithm design and delivers reusable technical pathways and practical guidelines for real-world deployment. (149 words)

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Segmentation
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Semantic segmentation is an important and popular research area in computer vision that focuses on classifying pixels in an image based on their semantics. However, supervised deep learning requires large amounts of data to train models and the process of labeling images pixel by pixel is time-consuming and laborious. This review aims to provide a first comprehensive and organized overview of the state-of-the-art research results on pseudo-label methods in the field of semi-supervised semantic segmentation, which we categorize from different perspectives and present specific methods for specific application areas. In addition, we explore the application of pseudo-label technology in medical and remote-sensing image segmentation. Finally, we also propose some feasible future research directions to address the existing challenges.

Problem

• Reduces need for large labeled datasets in semantic segmentation.
• Explores pseudo-label methods for semi-supervised learning.
• Focuses on medical and remote-sensing image segmentation applications.

Innovation

• Uses pseudo-labels for semi-supervised semantic segmentation
• Focuses on medical and remote-sensing image applications
• Proposes future research to overcome current challenges