VisiSign

Real-time Traffic Sign Detection System

A project developed in collaboration with Visimind – it uses the YOLO11 model for image analysis to detect traffic signs.

📚 Project Structure

VisiSign consists of three main components:

• 🖥️ Backend – FastAPI server with JWT authentication, WebSocket support, and real-time image processing using a pretrained YOLOv11 model
• 📱 Frontend – Mobile app (React Native, WIP) for capturing and sending images
• 🧠 Model – (Optional) Data pipeline and training code for retraining YOLOv11 on custom traffic sign datasets

---

🔄 Usage Scenarios

• ✅ For users/testers:
  You only need the Backend, Frontend, and a pretrained model file (e.g., VisiSign_Advanced.pt) — no training required.

• 🛠️ For developers/researchers:
  You can use the Model/ tools to download custom datasets, apply augmentations, train YOLOv11 from scratch, and log results to MLflow.

🧠 What does it do?

• Detects traffic signs from images sent via WebSocket
• Real-time processing using YOLO11
• Reports user results and statistics
• Secure authentication using JWT

🛠️ Technologies

• Frontend: React Native (Expo)
• Backend: Python + FastAPI + PostgreSQL
• ML: YOLO11 (Ultralytics)
• Experiment tracking: MLflow
• Docker: Fully container-ready

👨‍💻 BAKUSIE Team

Model:

• Artur Sierakowski

Backend:

• Filip Połom
• Grzegorz Jaworski

Frontend:

• Tomasz Budzejko

📄 Documentation

Documentation is available in the Docs folder

📄 License

This project is licensed under the MIT License.