Build a Teachable Machine Image Classifier With Your Kid

A hands on weekend project to demystify computer vision and machine learning.

Children today are surrounded by artificial intelligence, yet they rarely get to see behind the curtain. Most AI interactions are passive, such as chatting with a bot or watching recommended videos. To truly prepare the next generation for an AI driven world, we must help them transition from consumers to creators. This weekend project allows you and your child (ages 8 to 16) to build and train a custom image recognition model in fifteen minutes using a free tool developed by Google.

By using your computer webcam, your child will train a neural network to recognize different objects, hand gestures, or facial expressions. There is no coding required, no paid subscriptions, and no complex software to install. All you need is a web browser and a curious mind.

Step 1: Set Up Your Lab

To begin, sit down with your child at a computer equipped with a working webcam. Gather two or three distinct objects from around the house. Excellent choices include:

• Two different action figures or stuffed animals.
• A coffee mug versus a water bottle.
• Your hand showing a "thumbs up" versus a "peace sign."

Once you have your objects, open your web browser and navigate directly to the Google Teachable Machine website at https://teachablemachine.withgoogle.com/ . Click on "Get Started" and select "Image Project," then choose the "Standard Image Model."

Step 2: Gather Your Training Data

On the screen, you will see two default categories called "Class 1" and "Class 2." Think of these as the folders where your AI will store its visual memories. Have your child rename these classes to match your chosen objects (for example, "Stuffed Bear" and "Toy Car").

Now, click the "Webcam" button under the first class. Hold your first object up to the camera. Instruct your child to click and hold the "Hold to Record" button. As the camera captures images, tell your child to slowly rotate the object, move it closer and further away, and tilt it in different directions.

Aim to capture between 100 and 200 image samples. This variety teaches the computer what the object looks like from multiple angles and in different lighting conditions. Repeat this exact process for the second class using your second object.

Step 3: Train Your Model

Once both classes have ample image samples, look at the middle column labeled "Training." Click the "Train Model" button.

It is crucial to keep the browser tab open and active during this process. The training happens entirely inside your browser, meaning no data or webcam images are sent to Google servers. This is an excellent moment to explain to your child that the computer is currently looking at all the pixels in the images, finding patterns, and building a mathematical rulebook to tell the two objects apart.

Within thirty seconds, the training will complete, and the "Preview" column on the right side of the screen will activate.

Step 4: Test and Break the AI

Now comes the most engaging part of the project: testing the model. Have your child hold up one of the objects in front of the webcam. Watch the live confidence bars at the bottom of the preview screen. If the model is working correctly, the bar for that object should jump to 99 percent or 100 percent.

Next, encourage your child to try and trick the AI. Here are three ways to test its limits:

• The Halfway Test: What happens if you hold up both objects at the same time?
• The Angle Test: What happens if you turn the object completely upside down? If you did not train it on upside down images, the confidence score will likely drop.
• The Imposter Test: Hold up a completely different object (like your hand or a shoe). Which class does the AI think it is, and why?

Parent & Educator Sidebar

Core AI Concept: Supervised Learning This project demonstrates supervised learning. The computer learns by looking at labeled examples (the training data) provided by a human teacher. It does not actually "know" what a toy is; it simply recognizes mathematical patterns in the pixels.

Conversation Starters:

• "If we only trained the model with a blue toy car, why does it struggle to recognize a red toy car?" (Discusses dataset bias).
• "How do you think self driving cars use this technology when they look at traffic signs?"
• "Why is it important to have many different types of pictures when training an AI?"

Action Steps Summary

• Gather two distinct household objects to classify.
• Go to https://teachablemachine.withgoogle.com/ on a computer with a webcam.
• Record 150 webcam images for each object from various angles.
• Click "Train Model" and wait for the process to complete inside your browser.
• Test the model with your child and use the sidebar questions to discuss how AI learns.

Want to explore more practical, family friendly AI projects?