A deep learning assisted approach for plasterboard edge defects

Quality control methods used in gypsum board production are labor-intensive, time consuming processes that are prone to human error. Automatic defect detection systems using deep learning techniques offer a scalable solution to address such industrial challenges. This study aimed to develop a system for the automated detection of edge defects in gypsum boards using U-Net segmentation and classification models, implemented on a Raspberry Pi for industrial applications. The proposed system utilizes a U-Net model for defect segmentation and evaluates four deep learning models GoogleNet, AlexNet, ResNet-50, and Xception for defect classification. Performance metrics such as accuracy, sensitivity, specificity, and precision were analyzed. The system was developed using Python and deployed on a Raspberry Pi to assess real-time image capture performance.Training and validation metrics for the segmentation model were computed, while classification models were evaluated using confusion matrices and detailed metric comparisons. The U-Net model achieved a Dice Coefficient of 98.38% and a Mean Intersection over Union (IoU) of 96.29%, demonstrating robust segmentation capabilities. GoogleNet exhibited the highest overall accuracy (88.8%) among the classification models, with AlexNet excelling in sensitivity (96.6%) and ResNet-50 achieving the highest specificity (98.2%). This study highlights the effectiveness of deep learning techniques for automated defect detection in gypsum board production. The integration of segmentation and classification models with low-cost hardware such as the Raspberry Pi provides a scalable, efficientl in industrial environments. The findings underscore the potential of deep learning and edge computing to revolutionize defect detection in manufacturing processes.