🛡️ Aegis — AI-Powered Transit Monitoring

Aegis is an AI-powered transit monitoring app that enhances public safety by giving operators real-time situational awareness. It analyzes multiple live camera feeds, evaluates potential threats, and prioritizes the most critical streams so operators can respond faster and more confidently.

Key Features:

• Real-time threat detection: AI evaluates multiple camera feeds simultaneously.
• Prioritized alerts: Highlights high-risk streams to optimize operator response.
• Explainable AI: Provides clear reasoning behind flagged feeds.
• Responsive dashboard: Shows live feeds with actionable insights for operators.

Why We Built It:

Public transit safety is often reactive, with many incidents going unreported or addressed too slowly. Existing systems suffer from operator fatigue and inconsistent monitoring. Aegis closes this gap by providing an intelligent, operator-friendly system that enhances awareness without replacing human decision-making.

Technology Stack:

• Frontend: Vite + React
• Backend: FastAPI
• Streaming: WebSockets
• AI Processing: AWS Bedrock

Impact:

Aegis demonstrates how AI can improve transit safety, detect anomalies in real time, and help operators respond faster to critical incidents.

🎬 Demo

🌐 DevPost Project Page

---

🚀 How to Run

Set up both the backend and dashboard to get the app running locally.

🧠 Backend Setup

Prereq install ffmpeg:

Unix based OS (homebrew/linuxbrew):

brew install ffmpeg

Windows (Chocolatey):

choco install ffmpeg

Navigate to the backend directory:

cd backend

Create and activate a virtual environment:

python3 -m venv venv
# Unix based OS
source venv/bin/activate
# Windows
venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

Start the backend server:

uvicorn src.server:app --reload --port 3000

➡️ The API will be available at localhost:3000

💻 Dashboard Setup

Open a new terminal window.

Navigate to the dashboard directory:

cd HTC-dashboard

Install dependencies and start the development server:

npm install
npm run dev

➡️ The dashboard will be available at localhost:5173

🎥 Live Stream

To stream from another web client please go to: http://localhost:5173/webcam and start broadcasting the footage. Other clients will be able to view this footage live along with object and action detection from the footage.

🎥 Past Videos

To view old footage videos and analyze them in real-time please download the videos to the following directory: /backend/videos.

Example

/backend/videos/UnionStation.mp4