Introduction

Any-Cubes is an award-winning didactic toy for schools and maker spaces, with which children can intuitively and playfully explore and understand machine learning (image classification), as well as Internet of Things technology. It is a combination of deep learning-based image classification and machine-to-machine (m2m) communication via MQTT.

Using few-shot learning, children can intuitively train a visual classifier and integrate it into their play.

The Cubes

The system consists of three physical and tangible cubes:

Cube 1, the «Vision-Cube», can be trained to have a visual recognition of objects or scenes. Via the MQTT protocol, the Vision-Cube broadcasts its detected object / scene («1», «2» or «3») to other Any-Cubes devices («Light-Cube» and «Maker-Cube»).

Cube 2, the «Light-Cube», visualises the detected object / scene of the Vision-Cube by changing the color of its LEDs.

Cube 3, the «Maker-Cube»), can control external circuits using relays. It can, for example, be used to open an electronic cat door whenever the Vision-Cube detects a cat.

Apps

It addition to the physical cubes, apps can be created easily to react to the changes of the Vision-Cube. One such app is «Shapes»—it shows different graphical shapes to visualise the classes being detected by the Vision-Cube. It can be used as a boilerplate to create custom apps that react to the Vision-Cube.

Publications

Scheidt, A., & Pulver, T. (2019). Any-Cubes: A Children’s Toy for Learning AI: Enhanced Play with Deep Learning and MQTT. Proceedings of Mensch Und Computer 2019. https://doi.org/10.1145/3340764.3345375

Awards

• LEARNTEC Delina Award 2022: First place in the category «Frühkindliche Bildung und Schule» (Early childhood education and school)

Team / Contributors

Any-Cubes is a project by Alexander Scheidt and Tim Pulver with contributions by Meliani Meliani (research and organisation of workshops), Lukas Schmidt Wiegand (industrial design) and Sabina Fimbres Sabugal (graphic design for posters).

The project was heavily inspired by Google Creative Lab’s Teachable Machine and builds upon it.