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 in twenty minutes.

Most children understand that AI can generate text or images, but few understand how these systems actually learn. By building a custom image classifier, your child can transition from a passive consumer of AI to an active creator.

Teaching children aged 8 to 16 about artificial intelligence does not require complex coding environments or expensive subscriptions. Using Google's free, web-based tool, you and your child can build a fully functioning computer vision model. This hands-on project demonstrates how neural networks identify patterns in visual data, providing a foundational understanding of modern machine learning.

Step 1: Access the AI Lab

To begin, sit down with your child at a computer equipped with a standard webcam. Open a web browser and navigate directly to https://teachablemachine.withgoogle.com/ . Click on the "Get Started" button, then select "Image Project" followed by "Standard Image Model". This opens a clean, visual workspace divided into three main sections: Gathering Data, Training, and Previewing.

Step 2: Define Your Classes and Gather Training Data

An image classifier works by sorting visual inputs into distinct categories, which the platform calls "Classes". For this project, you will train the computer to recognize two different household objects or hand gestures (such as a thumbs-up versus a thumbs-down, or a coffee mug versus a pen).

First, rename "Class 1" to match your first object (for example, "Blue Pen"). Click the webcam icon within that class box. Instruct your child to hold the blue pen in front of the camera. Click and hold the "Record" button to capture at least one hundred images. Encourage your child to slowly rotate the pen, move it closer to the lens, and shift it to different areas of the screen. This variety helps the model learn what the pen looks like under different conditions.

Next, rename "Class 2" to match your second object (for example, "Red Apple"). Repeat the recording process, capturing another one hundred images of the second object from various angles and distances.

Step 3: Train Your Custom Model

With your training data collected, 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.

Explain to your child that the computer is currently analyzing the pixels in the images you provided. It is looking for patterns (such as colors, edges, and shapes) that consistently appear in the "Blue Pen" images but do not appear in the "Red Apple" images. This process typically takes less than one minute.

Step 4: Test and Debug the Classifier

Once training is complete, the "Preview" column on the right will activate, showing a live feed from your webcam. Have your child hold up the blue pen. The progress bars below the video should instantly shift to show a high confidence rating (close to 100%) for "Blue Pen". Now, have them hold up the red apple and observe the shift.

To make this a true learning experience, try to "trick" the model. What happens if you hold up a blue apple? What happens if you hold up a red pen? What happens if you block the light? When the model makes a mistake, explain that the AI is not actually "thinking" like a human, it is simply matching pixel patterns based on the limited data you provided.

Parent & Educator Sidebar

Core AI Concept: Supervised Learning and Bias This project demonstrates supervised learning, where an AI learns from labeled training data. If you only train the model with a blue pen held in your right hand, the model might mistakenly learn that "right hand" means "blue pen". This is a simple demonstration of algorithmic bias, showing how AI systems are limited by the quality and diversity of the data we feed them.

Conversation Starters:

• "Why do you think the computer got confused when we held up a different object that was the same color?"
• "How could we change our training images to make this model more accurate in a dark room?"
• "How might self-driving cars use this exact technology to recognize stop signs versus speed limit signs?"

Action Steps Summary:

• Go to https://teachablemachine.withgoogle.com/ on a computer with a webcam.
• Create two classes and record one hundred distinct images for each class.
• Click "Train Model" and observe the pattern recognition process.
• Test the model with different objects and discuss the concepts of training data and bias.

Want more family-friendly AI projects and executive workflows?