Image-Based Classification of Cutting Tool Degradation Using Deep Learning: A Comparative Study

Abstract

This study investigates the application of deep learning-based image classification techniques for cutting tool condition monitoring. Several models, including Convolutional Neural Networks (CNN), Visual Geometry Group 19-layer Network (VGG19), Vision Transformer (ViT), and EfficientNetV2, are evaluated using image data of cutting tools to classify their condition into good and defective. Transfer learning is employed to leverage pre-trained models and improve classification performance on the dataset. The models are compared based on evaluation metrics such as accuracy, precision, recall, and F1-score. Experimental results show that VGG19 achieves the best performance with an accuracy of 96%, followed by the CNN model with 93.6% accuracy. Vision Transformer achieves 87.5% accuracy, while EfficientNetV2 records 87.1% accuracy. The results indicate that VGG19 provides superior feature extraction and classification capability for the given dataset.