Build a Teachable Machine Image Classifier With Your Kid

Most children today are active consumers of artificial intelligence, whether they are interacting with voice assistants, playing games with AI-driven characters, or watching algorithmically curated videos. However, true digital literacy requires moving from consumption to creation. By building a functional machine learning model, children demystify the technology and realize that AI is simply a tool trained on data. Using a free, web-based tool, you and your child can build a custom image classifier in under thirty minutes. This hands-on project requires no coding experience, no paid subscriptions, and no specialized hardware, making it the perfect weekend activity to spark a lifetime of curiosity about computer science.

Demystifying Machine Learning with Teachable Machine

To build this project, we will use Google Teachable Machine, a web-based tool that allows anyone to train machine learning models quickly and easily. You can access the platform directly at https://teachablemachine.withgoogle.com/ . The platform uses a process called supervised learning, where we provide labeled examples of different objects, and the computer learns to recognize the patterns that distinguish them. For this activity, we will build a Rock, Paper, Scissors game helper that can instantly recognize which hand gesture your child is making in front of the webcam.

Step 1: Setting Up Your Classes

Once you open the Teachable Machine website, click on Get Started and select Image Project , then choose Standard Image Model . You will see a workspace with two default categories called Classes. Classes are simply the labels or categories that your model will learn to recognize. Since we are building a Rock, Paper, Scissors classifier, we need three classes:

• Rename Class 1 to Rock by clicking the pencil icon.
• Rename Class 2 to Paper .
• Click the Add a class button at the bottom and name Class 3 Scissors .

Step 2: Gathering the Training Data

Now comes the fun part: training the model. Your child will act as the data scientist, gathering the visual examples the computer needs to learn. For each class, follow these steps:

• Click the Webcam button inside the Rock class box.
• Have your child hold their hand in a fist (the Rock gesture) in front of the webcam.
• Click and hold the Hold to Record button. Capture about 100 to 150 images. While recording, have your child slowly rotate their hand, move it closer and further from the camera, and tilt it slightly. This teaches the computer to recognize a fist from different angles.
• Repeat this exact process for the Paper class (flat hand) and the Scissors class (two fingers extended). Ensure you capture a similar number of images for each class to keep the dataset balanced.

Step 3: Training the Model

With the data gathered, it is time to train the neural network. Click the blue Train Model button in the middle column. This process takes about ten to thirty seconds. During this time, the browser is analyzing the pixels in the images to find common features, such as the round shape of a fist or the long, separated lines of scissors. Make sure to keep the browser tab open and active while the training completes.

Step 4: Testing and Troubleshooting

Once trained, the Preview panel on the right will activate, showing a live feed from the webcam. Have your child make a Rock, Paper, or Scissors gesture. Below the video feed, you will see progress bars showing the model's confidence level in real time. If your child makes a fist, the Rock bar should jump to 100 percent.

If the model gets confused, this is a perfect learning opportunity. If it mistakes a hand gesture for something else, look at the background. Did the model accidentally learn to recognize your child's face or the chair behind them instead of the hand? To fix this, you can record more images with different backgrounds, or add a fourth class called Background where you record images of the empty room with no hand gestures. Train the model again and observe how the accuracy improves.

Parent and Educator Sidebar

The Core AI Concept: This project demonstrates Supervised Machine Learning . Computers do not inherently know what a hand gesture is. They learn by analyzing thousands of pixels in labeled training data to find mathematical patterns. If the training data is limited or biased, the model will make mistakes.

Conversation Starters:

• What happens if you try to use your left hand instead of your right hand? Why does the computer get confused, and how can we fix it?
• How do you think self-driving cars use this exact technology to recognize stop signs, pedestrians, and other vehicles?
• If we only trained this model with your hand, would it work well for a parent or a friend? Why is diversity in data important?

Action Steps Summary

• Access the Tool: Open https://teachablemachine.withgoogle.com/ on a computer with a webcam.
• Define Classes: Create three distinct classes labeled Rock, Paper, and Scissors.
• Record Data: Capture 100 to 150 webcam images for each hand gesture, rotating the hand for variety.
• Train and Test: Click Train Model, then test the live predictions in the preview window.
• Discuss: Use the sidebar questions to help your child connect this activity to real-world artificial intelligence systems.

Inspire the Next Generation of Innovators