Detection of Back Cover Material Defects Based on Convolutional Neural Network and TensorFlow

The application of deep learning using TensorFlow to improve the efficiency of material quality control is highly needed. In this case, the material quality inspection process in the form of a back cover is an important phase in the production chain. Damage resulting from industrial machinery reduces product quality. Some common problems occur when the back cover material is damaged, such as scratches, breaks, and cracks. This is caused by impacts or falls during the production process. It takes a long inspection time if done manually by workers one by one. The speed of material inspection is also an important priority for greater efficiency. An approach is proposed that applies deep learning using TensorFlow which focuses on convolutional neural networks (CNN) for image recognition and processing. It is used for image segmentation by detecting and separating objects. This method can classify the image quality of the back cover material. The main objective of this study is to produce a model that can detect material quality accurately. The highest level of accuracy was achieved at epoch 20 of 0.98. The success of this study is that it can reduce the involvement of inspectors. Thus, it can increase the efficiency of the defect detection process in the back cover material