AI Dentist Helper

The Project

This project is about an AI Dentist helper. It's an AI system that classifies dental X-ray images into categories like Implants, Fillings, Cavities, and Impacted Tooth to help dentists classify and diagnose faster and more accurately. My inspiration of this development was noticing that dentists were spending heavy hours on looking at hundreds of X-rays every day which is very time consuming and emotionally draining to many individuals, causing a lack of accuracy. I wanted to see if AI could make that job quicker and more consistent. I used a public Kaggle dental X-ray dataset and built everything in Google Colab. Instead of training everything from scratch, I used transfer learning with MobileNetV2, where I trained it for 10 epochs and added my own classification layers on top. The main struggle was fixing the dataset. When I downloaded it, the images were all in flat folders with an _annotations.csv file, so TensorFlow's image loader kept throwing errors and couldn't find any classes. I had to write extra code to reorganize the images into proper class folders. The dataset was also pretty small and imbalanced, so my final validation accuracy landed around 53%, as it was really good at spotting fillings but struggled more with cavities. Overall it was a lot of debugging, but I learned so much about real-world AI problems in healthcare.

View this project on YouTube