A dataset of images of human teeth, both with and without cavities, is gathered via Kaggle. From each of the three divisions of this data set, an equal number of images representing each category are selected for testing, training, and validation. The chosen images are then subjected to three preprocessing techniques: image resizing, image rescaling, and image augmentation. To develop a deep learning model, convolutional neural network technology is employed simultaneously. The model is then trained using the preprocessed data set. To assess the effectiveness of the training, accuracy and loss are employed. To understand how the model behaves, the values are also tabulated. The validation process follows the same phases. Analysis and plotting of the training and validation data into a bar graph. The results of the training and validation demonstrate that the model can analyze the image and predict the presence of dental cavities after only briefly evaluating it. A greater accuracy value and a very low loss value are produced by both training and validation. The model is then tested one last time after being trained and validated. The model's final accuracy is determined to be 96%, while the loss value is 11%. The accuracy and loss values of this deep learning model were found to be good, thus a mobile application was created using it. Additionally, the application's functionality is examined and found to be perfect.

Image Processing, Deep Learning, Data Collection, Accuracy, Loss