Build a Teachable Machine Image Classifier With Your Kid

Introduce your child to the fundamentals of machine learning by training a custom visual AI model using household items.

Children today are surrounded by artificial intelligence, from the recommendation algorithms on video platforms to the voice assistants in their homes. However, most kids remain passive consumers of these technologies, viewing AI as a form of digital magic. To prepare them for a technology driven future, they need to understand that AI is simply a tool that learns from patterns in data.

This weekend project allows you and your child (ideally ages 8 to 16) to build, train, and test a custom image classifier. Using a standard computer with a webcam and a free web browser tool, your child will step into the role of an AI engineer, gaining a practical understanding of how computers learn to see.

Demystifying Machine Learning

The best way to teach machine learning is through hands on experimentation. Instead of writing complex code, your child will use visual data to train a model. This approach mirrors the exact workflow used by professional data scientists: collecting data, labeling it, training the model, and testing its performance.

To begin, you do not need any paid subscriptions, specialized hardware, or prior programming experience. You only need a laptop or desktop computer with a working webcam and an internet connection.

Step 1: Choose Your Classification Project

Before opening the software, help your child decide what they want their AI to recognize. Choosing contrasting items makes the initial training easier and more rewarding. Here are three excellent starter ideas:

• Toy Classifier: Train the model to distinguish between a Lego brick and an action figure.
• Hand Gestures: Teach the computer to recognize a "Thumbs Up" versus a "Thumbs Down" or "Peace Sign."
• Fruit Detector: Use an apple and a banana to show how the computer distinguishes shapes and colors.

Step 2: Set Up Teachable Machine

Open your web browser and navigate directly to Google's Teachable Machine . Click on "Get Started" and select "Image Project," then choose the "Standard Image Model."

You will see a simple interface with three main columns: Classes, Training, and Preview.

• Rename the Classes: Change "Class 1" to the name of your first object (e.g., "Lego") and "Class 2" to your second object (e.g., "Action Figure").
• Record Training Data: Click the "Webcam" button under Class 1. Hold your first object in front of the camera. Click and hold the "Hold to Record" button. Move the object around slightly, showing it from different angles and distances. Aim for at least 100 image samples.
• Record the Second Class: Repeat this process for Class 2 using your second object, capturing a similar number of samples.

Step 3: Train and Test the Model

Once you have recorded data for both classes, click the "Train Model" button in the middle column. Instruct your child to leave the browser tab open and untouched while the computer processes the images. This step usually takes less than thirty seconds.

Once training is complete, the "Preview" column on the right will activate, showing a live feed from your webcam. Hold up one of the objects. The progress bars below the video will instantly show which object the AI thinks it is seeing, expressed as a percentage of confidence.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning This project demonstrates supervised learning, where an AI learns to recognize patterns by analyzing labeled examples (the training data). If the training data is limited or poor, the AI's predictions will be inaccurate.

Conversation Starters for Your Child:

• "What happens if we hold up an object the computer has never seen before? Why does it still try to guess?"
• "If we only train the model with a red Lego brick, will it recognize a blue Lego brick? Why or why not?"
• "How could a self driving car use this exact technology to keep people safe on the road?"

Introducing Edge Cases and Bias

To deepen the learning experience, challenge your child to break the model. What happens if you hold the Lego brick very close to the camera, or hide half of it behind your hand? What happens if you change the lighting in the room?

This is a perfect opportunity to explain algorithmic bias. If the AI was only trained on images of a Lego brick held in your hand, it might actually be learning to recognize your hand rather than the toy itself. To fix this, have your child record more diverse training data, including background shots without the objects, to help the AI isolate the correct features.

Action Steps Summary

• Gather Materials: Find two distinct household objects (e.g., a toy and a piece of fruit).
• Open the Tool: Go to Teachable Machine on a computer with a webcam.
• Record Samples: Capture at least 100 webcam images for each object, moving them to show different angles.
• Train the Model: Click "Train Model" and observe how the computer processes the visual data.
• Test and Refine: Use the live preview to test the model, introducing edge cases to see where it succeeds or fails.

Want to explore more family friendly AI projects? Share PromptHacker with other parents.