A new approach using hybrid pruning YOLOv5 and multiscale data augmentation has been proposed to enhance defect detection in substations. This method addresses challenges such as dynamic shooting environments and limited datasets, resulting in improved accuracy and stability.
