Main Content

In part one of a two-part project, we display thermal images on a TFT display, save them to flash, all to build an ML model w/ SMS alerts!

One thing they don’t tell you about raising chickens in the frozen tundra of Wisconsin, is how you’re routinely putting your life on the line just to go check for eggs.

As much as I’m willing to sacrifice my own children in the name of looking for an egg, I’d prefer to create an ML + IoT solution that can do it for me! And maybe save a child along the way.

In this two-part project, we will walk through how to create a Machine Learning-based IoT solution to detect the presence of a chicken egg using images provided by a thermal IR camera, relaying those inferences to the cloud using a cellular module, and sending an SMS notification if an egg is detected.

Part one (today) covers the hardware, wiring, and firmware needed to capture a thermal image, save it to an SD card as a PNG, and connecting to the cloud. Sounds simple, but it’s deceptively tricky!

Specifically, in part one we are going to cover:

- Using an STM32-based Swan MCU (but any modern MCU will do);
- Hooking the Swan up to a MLX90640 thermal IR camera;
- Displaying thermal images on a ILI9341 TFT display;
- Saving “screen captures” to an SD card (as PNGs) to use to train our ML model;
- Using a 7-segment display to show an active count of images;
- Relaying image counts to the cloud with a cellular system-on-module, the Blues Wireless Notecard.”

Link to article