Google's Creative Lab has officially released its TensorFlow.js-based in-browser machine learning system, the Teachable Machine, which aims to allow the training of machine learning models without any coding on the user's part — simply by gathering examples, training the system, and exporting for use in external projects.
First released back in 2017 as an experiment but now an official service, Google's Teachable Machine aims to bring the benefit of machine learning to more developers by simplifying the creation of machine learning models - to the point where no coding or previous machine-learning experience is required at all.
"Teachable Machine is a web-based tool that makes creating machine learning models fast, easy, and accessible to everyone," the company explains of its creation. "[It's for] educators, artists, students, innovators, makers of all kinds — really, anyone who has an idea they want to explore. No prerequisite machine learning knowledge required. You train a computer to recognise your images, sounds, and poses without writing any machine learning code. Then, use your model in your own projects, sites, apps, and more."
The current version of Teachable Machine supports three types of input — images and videos, sounds, and body positions - supplied either as previously-captured files or live streams from on-device sensors. For those concerned about privacy, Teachable Machine can also be run entirely on-device - meaning any input data remains local and is never transmitted.
To demonstrate its potential, Google has published a trio of tutorials — outlining the creation of models which can detect finger-snaps, claps, and whistles, the direction a person's head is tilting, or estimate the ripeness of a banana — along with a series of case studies on using Teachable Machine in projects including robotics, accessibility, and gaming.
Teachable Machine is available now on its dedicated website, while projects built with the tool can be found on the Experiments with Google site.