Google has launched a new online demo to help familiarize people with machine learning.
Called Teachable Machine, the demo is designed to let users train the system to recognize certain objects or behaviors using their desktop cameras, and to respond to these triggers with a GIF or audio. For example, a user could train the system to recognize when her hand’s fingers are positioned to resemble a dog’s face, and to then play the sound of a barking dog when this is detected. And it’s all done locally on the user’s device, helping to ensure privacy.
Google says the system generally needs at least 30 images to learn to recognize a particular visual cue, and warns users on its tutorial that their experiments aren’t always going to work exactly as they want them to: “Keep in mind that your machine doesn’t have an understanding of higher level concepts, like faces or objects,” the tutorial says. Nevertheless, the system could offer a fun way for users to get a better understanding of how machine learning works, and to appreciate what Google is up to with projects like its Google Lens system for mobile devices.
It could also help to spur further development in this area to everyone’s benefit, with Google having open sourced its deeplearn.js code to encourage further tinkering among developers.
Source: The Keyword