![](https://cdn.mos.cms.futurecdn.net/bPNxBUHqpUB7niGcmts4Gc.jpg)
We’re always excited to see the Raspberry Pi used in new and exciting ways—especially when those projects can help save lives. A maker known as Aneeshamol is using our favorite single-board computer to monitor for fires with AI.
The project relies on image recognition by monitoring a given area using a camera module. All you have to do is face it toward the area you want to monitor and train the AI to check for fires. Images from the video feed are evaluated by OpenCV, which scans each parsed frame. It checks for variations using the hue, saturation and value (HSV) color model. If a potential fire has been detected, it can automatically notify users using SMS.
![](https://cdn.mos.cms.futurecdn.net/ztmV5gTz5r6fqhqcYnje6a.jpg)
In the project page, Aneeshamol explains various use cases for using an image based AI system to monitor for fires. The system can be programmed to help identify smoke versus visually similar phenomena like clouds and help evaluate the severity of the situation. Aneeshamol also included a special board that adds mobile network support for notifications.
The board Aneeshamol selected to drive the fire monitor is a Raspberry Pi 4 B connected to a standard Raspberry Pi Camera Module. A Blue Notecard is attached with the help of a Blues Notecarrier-Pi. Although Aneeshamol is using a Pi 4, you could easily get away with using a Raspberry Pi 5 if you have one on hand.
Of course, this shouldn't be used in lieu of normal safety. Always have a fire plan, keep an extinguisher nearby (and know how to use it), and keep an eye on cooking, candles, campfires, and anything else with a flame.
Aneeshamol was kind enough to share all of the build details which includes a close look at the source code. A custom Python script is used and available for anyone who wants to check it out for themselves. It works alongside OpenCV to evaluate the HSV data using AI. The cellular network card is programmed using Blues Notehub.io. If you want to get a closer look at this Raspberry Pi project, visit the project page over at Hackster.