We focused on making it easier for people who are curious about machine learning to learn more about it. Teachable Machine is a simple experiment that lets you teach a machine, using your camera to respond to you in fun ways — live in the browser, no coding required.

Upon arriving on the Teachable Machine site, the user can use their webcam to train three separate classes on anything, such as facial expressions or hand gestures. The image classification used to train each class is powered by a neural network within the Tensorflow.js library. This library allows web developers to train and run machine learning models locally in the browser without the need for any large complicated machine learning servers. Once each class is trained, the user assigns a fun output, such as a cat GIF or a cow moo sound as a response to each class. When the user repeats the gesture for each respective class, the machine will respond with the assigned output. With Tensorflow.js, web developers are able to use javascript to train the machine and illicit responses in real time without complicated machine learning/engineering knowledge or big servers.

342,000 visitors from 203 countries, have spent 14,200 hours (595+ days) training their own machine. Teachable Machine is being utilized in various different college and high school level Computer Science, Machine Learning and Artificial Intelligence classes. For example, exploring Teachable Machine is the first assignment in Hal Abelson’s Intro to Machine Learning class at MIT. We also offer a Teachable Machine boilerplate github project which demonstrates how web developers can use Tensorflow.js to set up projects similar to Teachable Machine, something that up until now would require a high level of machine learning knowledge and engineering skills to complete.