Accurate and timely insulator defect detection is crucial for maintaining the reliability and safety of the power supply. However

the development of deep-learning-based insulator defect detection is hindered by the scarcity of comprehensive

high-quality datasets for insulator defects. To address this gap

the synthetic insulator defect imaging and annotation (SYNTHIDIA) system was proposed. SYNTHIDIA generates synthetic defect images in a 3D virtual environment using domain randomisation

offering a cost-effective and versatile solution for creating diverse and annotated data. Our dataset includes 22

000 images with accurate pixel-level and instance-level annotations

covering broken defect and drop defect types. Through rigorous experiments

SYNTHIDIA demonstrates strong generalisation capabilities to real-world data and provides valuable insights into the impact of various domain factors on model performance. The inclusion of 3D models further supports broader research initiatives. SYNTHIDIA addresses data insufficiency in insulator defect detection and enhances model performance in data-limited scenarios

contributing significantly to the advancement of power inspection.