<![CDATA[This study discusses the detection of medium voltage insulator cracks using object detection technology. This study uses medium voltage ceramic insulator image data at the ULP Daya Waste Material Warehouse. Ceramic insulator image data is categorized into good and damaged conditions. Preprocessing involves labeling, dividing train data, validation, and testing, and exporting data to Pascal VOC format. MobileNetV2 is implemented on Google Collab to train the object detection model. The evaluation of the model accuracy is in the COCO matrix, while the performance graph shows that the model can read objects well because the reading curve is in line with the smooth curve. Furthermore, this model is applied in creating an Android application that uses the device's camera to detect objects in real-time. This application processes images, converts from YUV to RGB, and performs object detection using the trained model. The detection results are displayed with bounding boxes and labels on the camera reviewer, namely the good class with a reading value of 1.0, the damaged class with a reading value of 0.67 and the background 1.0. This application also tracks detected objects and updates the display according to the detection results.]]>
Copyrights © 2025
