Automated power line inspection plays a crucial role in maintaining grid reliability within smart cities by identifying potential defects in towers, conductors, insulators, and fittings. While modern anomaly detection relies heavily on deep neural networks (DNNs), training these models requires massive amounts of high-quality image data. However, a significant scarcity of publicly available datasets persists because data acquisition not only demands highly specialized professional skills but also faces strict data protection regulations enforced by grid companies. To bridge this gap, this paper presents a comprehensive review of open-access image datasets dedicated to power line inspection. Based on strict inclusion criteria—specifically, unrestricted public availability and a direct focus on core power line components—19 datasets are systematically selected and analyzed. We provide a detailed taxonomy and comparative analysis of these datasets in terms of inspection targets, acquisition platforms, annotation toolkits, and labeling schemes. Furthermore, our investigation highlights current research trends and identifies critical gaps, such as the disproportionate focus on insulators and the notable scarcity of multimodal data. To address the limitations of small-scale datasets, we also discuss existing data augmentation strategies and synthetic data generation techniques. Ultimately, this review serves as a unified navigational guide, aiming to foster the development of more robust visual inspection algorithms and to inspire future high-quality dataset construction in the power domain.
Wu et al. (2026) studied this question.