Workshop: Getting to Know Google Teachable Machine

Background

Designers need to develop a first-hand understanding of the material properties of machine learning. This activity is about exploring this technology without the need to do any programming.

Learning Objectives

After this activity, you can create a machine learning model using Google Teachable Machine that classifies images.

Instructions

Step 1

  • Go to Google Teachable Machine: https://teachablemachine.withgoogle.com
  • Create a new image project
  • Upload the labeled images from the training dataset provided
  • Train the model
  • Preview the model by uploading a couple of pictures from the same dataset
  • Reflect on the results. How does the model perform? Is it what you expected? Why do you think it behaves as it does?

Step 2

If you have time, also do the following:

  • Re-use the training set from step 1, or collect new images yourself
  • Train your model
  • Upload the model to the cloud
  • Make a note of the URL
  • On your smartphone, go to https://kind.io.tudelft.nl/teachable-machine-viewer/
    • On iOS, make sure you use Chrome rather than Safari
  • Insert the URL for the model you just saved to the cloud
  • Walk around (go outside if you like) and point the phone camera at objects you think it should be classifying
  • Again, reflect on the results. How it performs, why it behaves the way it does, etc.

Product

The outcome of the activity is a trained model and some notes on your reflections.

Follow-up

We will discuss your reflections in the plenary session later in the day.