Predict
Students note which class they expect and why.
Smart Bin
Smart Bin has students train and test an image recognition model and connect its prediction to a microcontroller.
How it works
From photo to decision.
Test
The camera or a test image is evaluated by the AI model.
Inspect
The webpage shows the chosen class and the confidence for each class.
Connect
When the prediction is stable, the browser sends a code to the microcontroller.
Conclude
Students describe when the model works well and which examples are missing.
The setup
Make artificial intelligence explainable.
The class works with real images, a model from Teachable Machine and a microcontroller that responds to the prediction.
- The model recognises classes such as organic, plastic, metal and paper.
- The browser shows not only the answer, but also the confidence.
- The Arduino or micro:bit uses the received code to make the sorting action visible.
What will you learn?
An AI model is not a magic box.
Classes
Students link a visible object to a label the model knows.
Confidence
The confidence score helps discuss when a prediction is usable.
Dataset
Wrong predictions become a reason to look for missing or one-sided examples.
Serial communication
The browser sends a simple code to the microcontroller.
Testing
Students compare expectation, prediction and physical response.
- Model
- Image classes with Teachable Machine
- Output
- Prediction and confidence per class
- Practice
- Demo mode without camera or hardware
Place in the learning path
From platforms to projects.
The platforms build computational thinking. The projects use that thinking in a cross-curricular assignment in which students build, measure, test and conclude.
Platforms
STEaM projects
Smart Plants
Arduino, soil moisture, calibration and data.
Robotic hand
Arduino, sensors, servos and movement.
Wind
Micro:bit, energy, voltage and current.
Smart bin
AI, image recognition, data and microcontrollers.
Self-learning games
Getting started with machine learning
Particulate matter
Arduino, air quality, data and poster session.
For teachers
Scalable teaching material for primary education up to the second stage of secondary education.
This teaching material introduces students to AI technology. You choose how far to take it. Online only? Train your own AI model? Connect it to a micro:bit or build a full setup with your class?
- No student account required.
- WebSerial for Chrome and Edge.
- Demo mode when camera or hardware is unavailable.
- Report with name, class, prediction, observations and reflection.
Privacy and licence
Privacy
This project does not use its own analytics or advertising cookies. Input, progress and camera images are processed in the browser and may be stored locally on the device. The camera and connected hardware are used only after permission is given in the browser; the project does not upload the images. GitHub Pages logs visitors’ IP addresses for security purposes; external services are governed by their provider’s privacy policy.
License
Original project material by Robbe Wulgaert may be used, copied and adapted free of charge for education, provided that ‘Robbe Wulgaert – AI in de Klas’ is clearly displayed, where reasonably practical with a link to the original project. Commercial sale, paid distribution, sublicensing or repackaging as a commercial product is not permitted without prior written permission. Other uses outside education also require prior written permission. Third-party components remain subject to their own licences.