Build a Teachable Machine Image Classifier With Your Kid

Children today are surrounded by artificial intelligence, from video recommendation algorithms to voice assistants. However, most young people remain passive consumers of these technologies, viewing them as a form of digital magic. To prepare them for a future shaped by technology, we must help them transition from consumers to creators. By building a custom image classifier, children learn that AI is not magical, it is simply a system that recognizes patterns based on data we provide.

The Project: A Custom Object Detector

In this activity, you and your child will build a machine learning model that can instantly distinguish between different physical objects, such as a favorite toy, a book, or a hand gesture. This project requires no coding experience, no paid subscriptions, and no specialized hardware. All you need is a computer with a webcam and an internet connection.

We will use Google's free, web-based tool: Teachable Machine . This platform allows users to train machine learning models directly in their web browser, keeping all data private and local to the device.

Step-by-Step Build Guide

1. Choose Your Objects

Have your child select two distinct items from around the room. Good choices include a colorful mug, a stuffed animal, a gaming controller, or even different hand signs (like a peace sign versus a fist). The items should look noticeably different from one another to ensure early success.

2. Set Up the Classes

Navigate to Teachable Machine and click "Get Started," then select "Image Project" followed by "Standard Image Model." You will see two default boxes labeled "Class 1" and "Class 2." Have your child rename these boxes to match the selected objects (for example, "Stuffed Dinosaur" and "Blue Mug").

3. Gather Training Data

Click the "Webcam" button under the first class. Instruct your child to hold the first object in front of the camera. Click and hold the "Record" button to capture about 100 to 150 images.

Tip for success: Have your child slowly rotate the object, move it closer and further away, and tilt it. This teaches the AI to recognize the object from different angles, not just one static position.

Repeat this exact process for the second class using the second object.

4. Train the Model

Click the blue "Train Model" button. This process takes about ten seconds. Explain to your child that the computer is currently looking at all the pictures, finding patterns of colors, shapes, and edges that make the dinosaur different from the mug. Keep the browser tab open and active during this step.

5. Test and Debug

Once training is complete, a live preview window will appear on the right. Have your child hold up one of the objects. The bar charts below the preview will show the model's confidence level in real-time.

Try to "trick" the model. What happens if you hold the object upside down? What happens if you cover half of it with your hand? If the model gets confused, you can easily add more photos to the training classes and retrain it.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning This project demonstrates supervised learning, where an algorithm learns from labeled training data. The computer does not actually know what a "dinosaur" is: it only knows that the group of pixels labeled "Dinosaur" shares specific mathematical patterns that differ from the group labeled "Mug."

Conversation Starters:

• "How do you think the computer tells the difference between these two items? Is it looking at color, shape, or something else?"
• "If we trained the model using only your hand holding the mug, would it recognize the mug if I held it? Why or why not?"
• "How could a self-driving car use this exact same technology to keep people safe on the road?"

Action Steps Summary

• Gather two distinct household objects or decide on two different hand gestures.
• Open Teachable Machine on a computer with a webcam.
• Capture 100 to 150 images for each object class, rotating them to show different angles.
• Train the model and test its accuracy, discussing how data quality affects the results.

Inspire the next generation of builders.

Spend fifteen minutes this weekend demystifying technology for your child.

Make the activity stick

Have the child write down the examples before opening the tool. Three good examples for each label beat twenty rushed examples. If the classifier makes a strange prediction, treat that as the best part of the activity. Ask which examples confused it and what new example would make the boundary clearer.

Use Google Teachable Machine as the starting point because it lets kids see labels, examples, training, and testing in one browser window. No paid account is required for the basic activity, and the child gets a concrete model instead of a vague explanation.

Parent or educator prompt: ask what the model learned, what it failed to understand, and how a biased set of examples could change the result. The core AI concept is training data quality. Kids should leave knowing that models do not magically understand the world, they copy patterns from the examples people give them.