If a fire breaks out in your kitchen, garage, or warehouse while you're away, every second matters. This project adds a second pair of AI-powered eyes to whatever CCTV you already own, watching the live feed and sending an instant push alert to your phone the moment flames appear — and as long as the fire keeps burning, the alerts keep coming every 10 seconds so you don't miss it.
The whole thing is built with two visual blocks and two dashboard widgets — no code, no model training, just a drop-in fire detection layer for your existing camera.
- AI Fire Detection on any camera type — USB, CSI, ONVIF, or RTSP
- Real-time fire probability tracking
- Instant push notifications with tap-to-view live camera stream
- Repeating alerts every 10 seconds while the fire persists
- Three AI modes (Fast / Balanced / Accurate) for any device class
⚠️ Important safety note: This AI is designed to support your existing fire detection system, not replace it. Never rely on it alone for fire safety — always keep certified smoke alarms and fire detectors in place. Treat this as an extra alert channel, not a primary safeguard.
Why did we decide to make it?Most of us already have CCTV cameras pointed at the riskier areas of our homes and workshops — kitchens, boiler rooms, EV charging spots, server racks. Those cameras can record a fire, but they can't tell you it's happening. Adding a fire detection layer on top of footage you're already capturing is a tiny effort with a potentially huge payoff, so we built this as a reference project showing how a few visual blocks can turn passive surveillance into an active alert.
What You'll Build- An AI camera that recognizes flames in the live feed
- A live Fire Probability variable that climbs as flames appear
- A push notification trigger that fires when the probability crosses 50%
- A repeating alert loop that keeps pushing every 10 seconds while the fire continues
- A dashboard with the live camera and a real-time probability readout
- Download and install Grablo software on your device from grablo.co/download
- Get this project from the Grablo Gallery
- Configure your camera in Settings → Camera → Edit (every CCTV is different — see Hardware Setup below)
- Open the project, connect to your device, and hit RUN
- Install the Grablo app on your phone and sign in to receive push notifications
- Watch the Fire Probability climb when flames appear — push alerts land every 10 seconds while it stays above 50%
The dashboard has a Camera widget showing the live feed from your CCTV, and a Label widget bound to the Fire Probability variable with the unit set to percent so you can see the AI's confidence at a glance. The whole layout is duplicated to a mobile view so the same dashboard reads cleanly on your phone.
The project runs on two controls. The first one starts the AI: a Once condition fires it at launch, and an AI Analysis action creates a Fire Detection analyzer bound to your camera. An Add Analysis command tells the analyzer to write its detection result into a Fire Probability variable, and a Start Analysis command kicks off the model on the live feed.
The second control is the alert loop. A Compare condition watches the Fire Probability variable and triggers whenever it goes above 50, firing a Push Notification action with a fire alert title and message. A Delay action then pauses 10 seconds before the loop checks again — so as long as the probability stays high, the user gets a steady stream of alerts every 10 seconds.
Two controls for the AI and the alert, two widgets on the dashboard, and you have a working fire monitor.
Expected Results- The dashboard shows your live camera feed and a Fire Probability readout
- When flames appear in view, the probability climbs in real time
- Once it crosses 50%, a push alert lands on your phone
- While the fire persists, you keep getting a push every 10 seconds
- Tapping any notification opens the live camera stream so you can verify the situation immediately
This project is camera-agnostic on purpose — the project file ships with the camera connection deliberately blank because every CCTV is different. After importing, go to Settings → Camera → Edit and enter the connection details for your specific setup:
- Old analog CCTV without networking: connect it to your Grablo device through a USB video capture adapter, then choose CSI/USB type
- USB webcam: CSI/USB type, index 0 (or use the Scan button)
- Network IP camera: ONVIF or RTSP type with the camera's address and credentials
For the AI mode, pick Fast on low-power devices like a Raspberry Pi, Balanced for mid-range hardware, or Accurate when you want the best detection on a more capable machine.
To receive push notifications, install the Grablo app on your phone and sign in with the same account — the alerts route automatically to every signed-in device.
What's Next?- Lower the 50% threshold for more sensitive areas, or raise it where false positives are common (welding spaces, kitchens with open flames)
- Shorten the 10-second delay for higher-risk zones, or lengthen it to reduce notification fatigue
- Add a Send Email action alongside the push for a backup notification channel
- Trigger a smart plug or relay to cut power to the room when fire is detected
- Combine multiple cameras (kitchen, garage, server room) into one fire-monitoring dashboard
- Camera feed not showing: Double-check Settings → Camera → Edit — the project ships without camera credentials, so you need to enter your CCTV's address or USB index manually
- Fire Probability stays at 0: Make sure the AI Analyzer started successfully (check the project log), and confirm the camera feed is visible — the AI only runs when the camera is delivering frames
- No push alerts arriving: Install the Grablo app on your phone, sign in with the same account, and allow notifications. Verify the Compare condition is set to Greater than 50 and the variable is bound correctly
- Too many false alarms: Raise the 50% threshold, or switch the AI mode to Accurate for fewer false positives at the cost of more compute
- AI is slow or laggy: Switch the AI mode to Fast
Comments