SentinelAI

• 

  Realtime camera feed with no detected object

• 

  Test image containing fire and smoke displayed alongside the dashboard interface

• 

  Fire-only test image used for detection validation

• 

  Combined fire and smoke detection test image

• 

  Fire and smoke detection sample image

• 

  Dashboard showing a full log of past and recent detection events

• 

  Viewing individual detection events using randomly generated event IDs

• 

  Another view of a randomly selected ID detection event

• 

  Expanded event view with additional detection details displayed at the bottom of the dashboard

• 

  Telegram alert containing detection details and captured image

• 

  Telegram notification showing real-time fire/smoke alert

• 

  Email inbox showing unread fire detection alerts

• 

  Opened email displaying full detection details

• 

  WhatsApp alert showing detection notification with text-only event details

Inspiration

Fire outbreaks and smoke incidents often go unnoticed in their early stages, especially in homes, offices, warehouses, and industrial environments without intelligent monitoring systems. We wanted to build a smart AI-powered surveillance solution capable of detecting smoke and fire in real time and instantly notifying users before the situation escalates.

What it does

SentinelAI is an AI-powered CCTV surveillance system that detects smoke and fire in real time using computer vision. Once detection is confirmed, the system captures evidence, stores the incident, and instantly sends alerts through WhatsApp, Telegram, and email.

The platform also includes a smart dashboard for monitoring incidents, viewing event history, and reviewing captured evidence.

How We built it

We built SentinelAI using Python, OpenCV, YOLO-based computer vision models, Gradio, and alert integration services.

The system works by:

• Capturing live camera footage
• Running real-time smoke/fire detection
• Triggering alerts after stable detection
• Capturing screenshots and event evidence
• Saving incident history into a storage system
• Displaying incidents through a CCTV-style dashboard

We also implemented cooldown logic and event locking to reduce duplicate alerts and improve system stability.

Challenges We ran into

One of the biggest challenges was handling real-time video processing while simultaneously sending alerts and storing event data without freezing the system.

We also faced challenges with:

• Real-time detection stability
• Preventing duplicate alerts
• Integrating Telegram, WhatsApp, and email notifications together
• Managing performance on CPU-only systems

Accomplishments that We are proud of

• Successfully built a real-time AI smoke and fire detection system
• Integrated multi-platform instant alerts
• Developed a CCTV-style monitoring workflow
• Built an event storage and monitoring dashboard
• Created a scalable architecture suitable for future smart surveillance systems

What We learned

Through this project, we learned more about:

• Real-time computer vision pipelines
• Event-driven AI systems
• Alert automation workflows
• CCTV-style surveillance architecture
• System optimization and asynchronous processing

What's next for SentinelAI

Future improvements include:

• Cloud deployment
• Multi-camera support
• Live dashboard streaming
• AI-based false alarm reduction
• Edge deployment on embedded hardware
• Smart emergency escalation systems
• Mobile application integration

Built With

• ai
• api
• bot
• computer
• gradio
• learning
• machine
• opencv
• python
• smtp
• telegram
• twilio
• ultralytics
• vision
• yolo