Inspection System for Glass Bottle Defect Classification based on Deep Neural Network

The problem of defects in glass bottles is a significant issue bottle manufacturing. There are various types that can occur, including cracks, scratches, and blisters. Detecting these crucial for ensuring the quality production. inspection system must be able to accurately detect automatically determine affect its appearance functionality. Defective identified removed from production line maintain product quality. This paper proposed defect classification using Convolutional Neural Network with Long Short-Term Memory (CNNLSTM) instant base classification. CNNLSTM used feature extraction create representation class data. predicts anomalies based on similarity representations convolutional layer method incorporates transfer learning algorithm, pre-trained models such as ResNet50, AlexNet, MobileNetV3, VGG16. In this experiment, results were compared VGG16, ADA, Image threshold, Edge detection methods. experimental demonstrate effectiveness method, achieving high accuracies 77% body dataset, 95% neck an impressive 98% rotating dataset.