This CircuitPython-Powered Smart Lamp Uses TinyML to Spot Faces, Match Colors

Mononymous human-centered engineering student Leonard has put together a guide to building a CircuitPython-powered desk lamp with a difference: it features on-device artificial intelligence (AI) to vary its output when it detects a face.

"This project is the final project for a class at Boston College called 'Physical Computing' with Prof. [John] Gallaugher," Leonard explains of his creation. "In this project, you will be making a smart lamp using [a DFRobot] Huskylens AI and a Raspberry Pi Pico running CircuitPython."

The 3D-printed lamp uses three key components: a Raspberry Pi Pico RP2040-based microcontroller development board, running the CircuitPython firmware, a DFRobot Huskylens smart vision sensor based on the Kendryte K210 system-on-chip with integrated neural processing unit (NPU), and a strip of Adafruit NeoPixel RGB LEDs. The idea: a lamp that can alter its colors based on the output of a tiny machine learning (tinyML) model.

Once assembled and flashed with Leonard's CircuitPython program, the Huskylens starts searching for faces through a hole in the lamp's lid. Once a face is detected using a computer vision algorithm running on the Huskylens, an idle animation on the RGB LEDs is interrupted for a smooth fade to white — then swapped out for a color-block classification algorithm, which fades between detected colors.

The project is documented in full, complete with source code, on Leonard's Instructables page.