Fault Detection of Electric Motor Coil by YOLOv3 with Spatial Attention

Abstract

Object detection has been widely applied to the visual inspection of factory products. Moreover, because the detection model must be improved based on the object and problem set, the model parameters must be fine-tuned and new feature extractors must be introduced. We present an automatic fault detection method for electric motor coils based on deep learning in this manuscript. To the best of our knowledge, this is the first deep learning approach for fault detection of electric motor coil. Furthermore, we combine the spatial attention mechanism with the object detection method YOLOv3 to highlight the location information of the defective part in the image. We built a real-time detection system so that anyone could use the detection model we formed.