IMAGE DETECTION OF DENTAL DISEASES BASED ON DEEP TRANSFER LEARNING

Abstract

Traditional dental disease detection is done by doctors using naked eyes directly, which contains many uncertain factors for misdiagnosis and missed diagnosis. In order to improve the accuracy and efficiency of the detection of dental diseases, a dental disease image detection assistance system based on deep transfer learning is designed, which can autonomously recognize the photos obtained from the camera that assists the doctor in the detection. Performing transfer training on the trained model on the tooth data set, retain all pretrained convolutional layer parameters, and fine-tune the model to be more suitable for tooth image recognition. At the same time, AlexNet, GoogLeNet, and VGG models will be used for traditional deep learning training and the results obtained will be compared and analyzed with the results obtained by deep transfer learning in terms of accuracy and timeliness.