Ontology-Integrated Machine Learning in Computer Vision: A Survey

Abstract ―​

Computer vision systems leverage machine learning techniques to interpret visual data and extract valuable information for a wide range of applications, including object and facial detection, pattern and facial recognition, image classification and annotation, and scene understanding. The integration of ontology with machine learning approaches has emerged as a promising strategy to enhance the interpretability, semantic knowledge, and performance of computer vision systems. This survey offers a comprehensive overview of recent advancements in ontology-integrated machine learning methods for computer vision tasks. It delves into the principles behind ontology-driven approaches, reviews state-of-the-art techniques and methodologies, and analyzes their effectiveness in improving the performance of computer vision systems. It also provides a systematic literature review of 26 high-quality articles recently published from 2011 to 2024, highlighting the diverse ways in which ontology is integrated with machine learning to enhance accuracy and interpretability across various domains of computer vision. Additionally, it emphasizes the superiority of interdisciplinary domain learning-based methods in handling large, complex datasets, offering end-to-end learning capabilities, adaptability, robustness, continuous advancements, and scalability.