<![CDATA[This study aims to develop an automated system for inspecting and sorting defects in paint cans based on the integration of deep learning and a conveyor belt to improve the efficiency and consistency of quality control in the industry. The methods used include the design of a mechanical conveyor system, the integration of electronic circuits, and the development of an object detection model using YOLOv12 with a multi-camera configuration to minimize blind spots. The dataset consists of 1,209 images divided into training, validation, and test sets, with data augmentation applied to improve model robustness. Evaluation was conducted using precision, recall, F1-score, and mAP metrics, along with end-to-end system testing based on sorting accuracy, latency, and throughput under various lighting conditions and conveyor speeds. The research results show that the model achieved a precision of 0.98, a recall of 0.96, an F1-score of 0.97, and an mAP50–95 of 0.96. However, the system implementation yielded a sorting accuracy of 55.6% with optimal performance at moderate speeds, indicating a significant influence of operational factors on the system’s overall performance.]]>
Copyrights © 2026
