Automating Powerline Inspection Data Analysis using Artificial Intelligence

Insulator defect detection is a critical task in powerline maintenance, essential for ensuring the reliability and safety of power lines. This paper presents ongoing research that aims to address the low performance to identify small objects or defects by generating a balanced image dataset with diverse defects using DALL-E and improving the accuracy of insulator defect detection in powerline infrastructure. The study builds on a transfer learning from Detectron2 model framework with a Faster R-CNN backbone, trained on 1280 labeled images of powerline insulators. Ongoing efforts focus on dataset enhancement and cross-validation of the performance.