How arbaaz-khan-tech/hackverse_2025_suraksha_setu works
Project type and purpose
• SurakshaSetu is a full-stack, AI-powered public-safety platform.
• It comprises three major subsystems:
• AI Detection Modules (facial recognition, hidden-weapon detection, intrusion sensing)
• Police Dashboard (Flask web app)
• Citizen Mobile App (Expo React Native)
File structure
• Top-level directories:
• Facial_Recognition(Criminal+Missing Person Detection) – Flask app + embeddings
• Hidden_Weapon_Detection – Streamlit app + ML weights
• The_Sword_App – renamed “backend” folder is actually the Police Dashboard, plus a “citizen” folder for React Native
• Config & code:
• app.py in each subfolder – entry points for each service
• embeddings/*.npy – stored face vectors
• *.pt files – YOLO/Infrared weapon-detection models
• requirements.txt – Python deps
• citizen/src – mobile screens, socket service, API config
• backend/templates & static – Jinja2 pages, JavaScript libs
README for additional insights
• AI modules run as independent microservices or local processes.
• Police Dashboard (Flask) calls ML services and MongoDB.
• Citizen App communicates via REST/Socket to the Flask backend.
• Uses Twilio for SMS/WhatsApp, MongoDB Atlas for data persistence, NVIDIA GPU for inference.