AI 231 Final Project
Role
Technologies
Repository
View on GitHub→Overview
Early detection and accurate classification of brain tumors is critical for effective medical treatment, but traditional diagnostic methods can be time-consuming and require specialized expertise. This project addresses the challenge of brain tumor classification from medical imaging using deep learning techniques. The system employs transfer learning by fine-tuning a pre-trained convolutional neural network (CNN), specifically removing the classification head and retraining the final layer for brain tumor image classification.
Implementation
The project is architected as a full-stack application with three main components: a Svelte.js frontend for user interaction, a FastAPI backend for request handling and business logic, and a dedicated ML microservice for model inference. The ML microservice uses PyTorch to load and serve the fine-tuned CNN model, which was trained by removing the head of a pre-trained network (such as ResNet or VGG) and replacing it with a custom classification layer for brain tumor categories. The entire system is containerized using Docker, ensuring consistent deployment across environments, with PostgreSQL serving as the database for storing user data, predictions, and metadata. The microservices architecture allows for independent scaling and deployment of the ML inference service, while the FastAPI backend handles API routing, authentication, and data persistence.
