Build a Teachable Machine Image Classifier With Your Kid

Help your child build, train, and test their first computer vision model in under twenty minutes.

Most children interact with artificial intelligence as passive consumers. They watch recommended videos, ask voice assistants to play music, or use filters on social media.

To prepare the next generation for an AI-driven world, we must shift them from consumers to creators. Demystifying how computers learn does not require complex coding or expensive software. Using Google's free Teachable Machine platform, children aged eight to sixteen can build a fully functional computer vision model. This hands-on project allows them to train a neural network to recognize physical objects, hand gestures, or facial expressions, providing an intuitive understanding of how modern AI systems perceive the physical world.

Step 1: Set Up Your AI Laboratory

To begin, you only need a computer with a working webcam and an internet connection. No account creation, software installation, or payment is required.

Open your web browser and navigate directly to https://teachablemachine.withgoogle.com/ . Click on the "Get Started" button, and select "Image Project" followed by "Standard Image Model".

Before you start clicking, decide together what your model will classify. A classic, highly engaging option is a "Healthy vs. Unhealthy Snack" classifier, or a "Rock, Paper, Scissors" game helper. For this guide, we will build a classifier that distinguishes between a coffee mug and a writing pen.

Step 2: Gather Your Training Data

An AI model is only as good as the data used to train it. On the screen, you will see two default classes: "Class 1" and "Class 2".

• Rename the Classes: Click the pencil icon next to "Class 1" and rename it "Mug". Rename "Class 2" to "Pen".
• Capture Mug Images: Click the "Webcam" button under the Mug class. Hold your coffee mug in front of the camera. Click and hold the "Hold to Record" button. Have your child slowly rotate the mug, move it closer to the camera, and then further away. Capture at least one hundred images. This teaches the AI what a mug looks like from different angles.
• Capture Pen Images: Repeat this exact process for the Pen class. Ensure you capture a similar number of images (around one hundred) to keep the dataset balanced. Use different pens if available to make the model more robust.

Step 3: Train Your Neural Network

Once you have gathered your training data, look at the middle column labeled "Training". Click the blue "Train Model" button.

It is crucial to leave the browser tab open and untouched during this process. The training happens directly inside your browser, using your computer's local processing power. Explain to your child that the computer is currently analyzing the pixels in the images, looking for patterns (like curves for the mug and straight lines for the pen) that distinguish one class from the other. This process typically takes less than thirty seconds.

Step 4: Test and Hack the Model

Once training is complete, the "Preview" column on the right will activate, showing a live feed from your webcam. Below the video feed, you will see two progress bars representing the confidence score (0% to 100%) for each class.

Have your child hold up the mug. The "Mug" bar should jump to nearly one hundred percent. Now, hold up the pen. The "Pen" bar should dominate.

Now comes the most educational part: trying to "trick" the AI. What happens if you hold up a thermos? What happens if you hold up a pencil? What happens if you block half of the mug with your hand? This experimentation helps children understand that the AI does not actually know what a "mug" is. It is simply matching pixel patterns based on the limited data it was given.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning

Supervised learning is the process of training an AI model using labeled data (in this case, images labeled "Mug" or "Pen"). The computer learns by finding mathematical patterns in the training data and applying those patterns to new, unseen data.

Conversation Starters:

• "Why do you think the computer got confused when we held up a spoon? How could we update our training data to fix that?"
• "If we only trained the model with blue pens, would it recognize a red pen? What does this tell us about bias in AI?"
• "How do you think self-driving cars use this exact technology to recognize stop signs and pedestrians?"

Action Steps Summary

• Open the Tool: Go to https://teachablemachine.withgoogle.com/ and start an Image Project.
• Collect Data: Capture one hundred webcam images for two different physical objects.
• Train the Model: Click "Train Model" and observe how the browser processes the images.
• Test and Discuss: Try to trick the model with new objects and discuss how the AI makes decisions.

Looking for more weekend AI projects to build with your family?