Detecção de Isoladores em Redes de Distribuição utilizando Aprendizado Profundo

The present work presents a methodology for image detection of four types of insulators used in medium voltage overhead distribution networks. For this, the OPDL dataset, which has images of intact and defective insulators in external and internal (laboratory) environments, was used as input for detectors based on Deep Learning. Different variations of the You Look At Once Version 5 (YOLOV5) architecture were used and evaluated considering detection accuracy and processing time. The obtained results show that the proposed detector offers high accuracy and sort processing time.