Skip to content

🚀 AIC-HCMUS Fragment Segmentation

Welcome to AIC-HCMUS Fragment Segmentation—your all-in-one solution for analyzing and reconstructing document fragments and broken objects. Powered by cutting-edge computer vision and deep learning, our platform delivers precise, reliable results for fragment segmentation and equivalent diameter estimation.

✨ Features at a Glance

🖥️ Frontend

  • Modern UI built with React & TypeScript
  • Stylish layouts using TailwindCSS
  • Seamless navigation via React Router
  • Secure user authentication (login/register)
  • Effortless image upload, prediction, and result visualization (overlaid masks, equivalent diameter data)

⚡ Backend

  • Fast, scalable API with FastAPI
  • JWT-based authentication
  • Endpoints for uploading images, running predictions, and retrieving results
  • Integrated with YOLOv11m for segmentation & equivalent diameter calculation
  • Data stored in PostgreSQL; images & masks in MinIO

🤖 Machine Learning

  • State-of-the-art YOLOv11m segmentation model
  • Automatic calibration object detection (e.g., red balls) for accurate equivalent diameter estimation
  • Generates overlaid masks and computes object equivalent diameters

☁️ Infrastructure

  • Docker Compose for easy local setup
  • Kubernetes manifests for scalable deployment (PostgreSQL, MinIO, app, NGINX)
  • Automated CI/CD with GitHub Actions—deploys straight to Google Kubernetes Engine (GKE)

🗺️ System Architecture

System Architecture Diagram