Comparative Performances of the Convolutional Neural Network based Transfer Learning Models for Classification of Dental Disease

The prevalence of large-scale dental disease in humans affects their health by making them unable to perform basic tasks with their mouth. Around 3.5 billion people worldwide are affected by dental and oral diseases, mostly in low- and middle-income countries. Due to the high cost of such a diagnosis and treatment, there are no adequate dental and oral health services. Utilizing rapid and dependable methods for early diagnosis of dental disease will be extremely beneficial in resolving this issue. An active area of research is identifying dental diseases from digital camera images. Various machine learning algorithms and the use of convolutional neural networks (CNN) have improved the accuracy of dental disease classification. A dental disease detection and classification system based on computer vision and dental disease photos will be crucial in this setting. Although a number of methods are currently used on an ad hoc basis to identify and classify dental diseases, there have been few systematic studies done to examine how well these methods work with real-world data on dental diseases. This study aim to assess the exhibitions of different CNN-based move learning structures, for example, the ResNet50V2, EfficientNetV0B0 and The Oral Disease Dataset, the Oral Infection Dataset, and the Teeth Dataset were the three publicly accessible databases on which the models were tested. Metrics performances such as, Accuracy, Review, F1 score and in general exactness were utilized to assess the end-product of the examination, which uncovered that ResNet50V2 showed the most elevated precision of 82%, 82.5% and 81.6% on the three datasets, separately, beating any remaining cutting edge models like EfficientNetV1B0 and EfficientNetV2B0.