Improving Quality Assurance: Automated Defect Detection in Soap Bar Packaging Using YOLO-V5

In this study, a deep learning algorithm based on the You Only Look Once (YOLO-V5) approach is proposed for the visual inspection of soap bar packaging defects. The defects considered in this study include pack missing, pack opening, and label errors. Images capturing various packaging defects were collected using a hardware setup consisting of a USB camera and a laptop. The collected image dataset was processed using the annotation tool Labellmg. Subsequently, the defect detection model was trained using YOLO-V5 and tested on different images as well as on live camera feeds. An Arduino Mega connected to the laptop via serial communication (USB cable) was utilized to trigger an LED indicating the detection of a defect. The proposed research achieved an accuracy of 98% with an average processing speed of the trained model is 200ms per input image or frame.