Skip to content

coskyler/response-one

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 

Repository files navigation

ResponseOne

ResponseOne is an AI-powered incident detection and response planning system that monitors thousands of live CCTV camera feeds in real time.

The system analyzes footage across Florida, detects and classifies potential incidents using Gemini, displays incidents on a map, and generates suggested response routes and plans.

Features

  • Real-time CCTV footage analysis
  • Incident detection and categorization using Gemini
  • Map-based incident visualization
  • Suggested response route generation
  • AI-assisted response plan generation
  • Parallel video processing for scalable stream analysis

Tech Stack

  • Frontend/API: Next.js, TypeScript
  • Footage Processing: Python, Gemini API
  • Database: MongoDB

Architecture

ResponseOne is split into two main components:

  1. Web Application

    • Built with Next.js
    • Provides the user interface and API routes
    • Fetches incident data from MongoDB
    • Displays detected incidents and response recommendations
  2. Processing Server

    • Continuously monitors and processes CCTV feeds
    • Sends frames and context to Gemini for analysis
    • Records detected incidents in MongoDB for use by the web application

How It Works

  1. Live CCTV feeds are continuously monitored.
  2. The processing server analyzes footage in parallel.
  3. Gemini evaluates image and text context to detect possible incidents.
  4. Detected incidents are categorized and stored in MongoDB.
  5. Users can view incidents, generate routes, and receive suggested response plans.

Challenges

Processing live CCTV feeds at scale requires careful handling of concurrency, reliability, and cost. Low-quality footage also creates model accuracy challenges, especially around false positives. The system prioritizes minimizing false negatives, since missing a real incident is more costly than flagging a questionable one.

Future Improvements

  • Improve detection accuracy through model and prompt testing
  • Expand camera coverage
  • Improve routing and response-plan generation
  • Add severity scoring and incident confidence levels
  • Add operator review workflows

About

Monitors thousands of CCTV cameras in real time with AI to detect and classify incidents before they're reported.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Contributors

Languages