Insulator defect detection using autonomous aerial vehicles (AAVs) images is a promising method for power transmission line inspections.However, varying sizes, orientations, and complex backgrounds of insulator defects result in high false negatives and low accuracy.Previous Cane Accessories studies have not adequately incorporated self-attention mechanisms focusing on adjacent keys.To address this, we propose an improved YOLOv8-based detection algorithm.We added a Contextual Transformer module to the YOLOv8 backbone for better contextual understanding and introduced a Partial Convolution layer to reduce redundant calculations.
Our model shows improvements over existing ones, achieving a precision of 97.5%, a mean average precision of 86.2%, and a recall of 81.1%, offering a robust solution for Feminine Hygiene Products automated, precise power line inspections.