🔫 AutoSentinel: Gun Detection with a Web Dashboard

🧠 Overview

AutoSentinel combines server-side video processing with web-based visualization tools to deliver a near real-time firearm detection solution for schools and secure facilities (It uploads frames, so it has a detection delay of ~3 seconds). The system ingests video feeds, runs detections using a high-performance computer vision model (Yolov11m), and broadcasts detected images from a surveilance stream to a web dashboard with actionable timestamps of where any guns were detected.

---

Multi-User Support

User Accounts & Authentication

• Secure authentication system with hashed passwords.
• Admins can register new users, assign roles, and manage system access.

Access Control

• Feeds and detection logs are only accessible to authorized users.
• Access permissions are enforced at the database.

Role-Based Permissions (Soon to be implemented)

• Admins: Manage camera feeds, user accounts, and system-wide settings.
• Guests: Read-only access to specific feeds or events.

---

Multi-Camera Feed Management

Camera Registration

• Camera streams can be registered and configured via the surveilance page for a respective dashboard (any detected cameras will automatically be used).

Stream Monitoring Dashboard

• Each feed is displayed in a dedicated dashboard tile.
• Users can:
  • View recent detection snapshots and status
  • Filter feeds by location, detection time, or returned Yolov11 confidence scale (Soon to be implemented).

Tech Stack

Backend (Server-Side)

• Node.js + Express.js

  • Core REST API layer for detection events, logs, and camera metadata.
  • Structured MVC-style service separation for clean logic management.

• Custom Video Processor Service (Python)

  • Script that runs firearm detection on video streams using a YOLOv11 model.
  • Communicates with Node.js via child processes for logging and event handling.

• PostgreSQL

  • Stores timestamped detection logs, metadata, and system events.
  • Optimized schema for querying recent detections and historical patterns.

Frontend (Dashboard)

• React.js + Vite

  • Fast, lightweight SPA with detection feed, camera stream previews, and historical data browsing.
  • Fetches from Postgres via repeated Tanstack query.

• Material UI (Soon to be implemented)

  • Utility-first styling for a responsive and accessible UI.

Infrastructure & DevOps (Soon to be implemented)

• **Docker
  • Entire system is containerized: web, API, video processor, and database.
  • Easy local spin-up and seamless transition to production deployment.

---

Screenshots