Dichroic LED Hand Control

Using machine learning to track user's hands, mapping hand position to LEDs on strips surrounding and illuminating dichroic tiles allowing the user to control the light and art that they see.

An interactive light sculpture where the colors are generated by the viewer’s hand movements.

The sculpture is made of pvc, wood, dichroic tiles, led strips, a teensy micro controller, and a webcam.

The leds are positioned in a circle around the dichroic tiles which are illuminated by the light traveling around in a circle creating a rainbow pattern on the face of the sculpture.

When you step in front of the webcam and raise your hands, the moving light stops and can be controlled with your hand movements.

This is accomplished through a machine learning algorithm running in a P5.js sketch that identifies hands. The center of your hand is mapped to the led strips allowing you to control the colors produced by the light sculpture.

After user testing, I decided that the piece works better with a potentiometer to control the speed of the lights. Users were not able to understand that they needed to have their hands far away from the pieces so that the webcam could register them. Additionally, the webcam / machine learning algorithm has difficulty finding viewer’s hands in the dark. It would likely work much better with a Kinect.

The final version of Dichroic Incidence: Two Hands. Two identical dichroic grids, one using Dichroic Blaze, the other Dichroic Chill tiles illuminated by LED strips surrounding the tiles. A knob on the side of the piece allows you to control the speed of the light turning the illuminated area from almost a video to stop motion.
DichroicIncidencePinkYellow.png
DichroicIncidenceBluePurple.png
DichroicIncidence2Circles.png

Making the sculpture

Connecting machine learning hand tracking algorithm using a laptop webcam in a P5.js sketch to serial output to a Teensy microcontroller mapping hand position to specific leds in a strip surrounding a grid of dichroic tiles.