Google’s AIY Projects (a clever play on “DIY” and “AI” that makes no sense when you actually think about it) pair a Raspberry Pi with the accessories, software, and requisite cardboard to make your own Google Assistant smart speaker or object-recognizing smart camera. They only launched last year, but Google is already back with new and improved versions of both the AIY Vision Kit and AIY Voice Kit (as spotted by 9to5 Google).
The new kits now include a fresh Raspberry Pi Zero in the box and a preloaded SD card to make it easier to get started. Google is also now providing an AIY companion app for Android (and soon iOS and Chrome) to help with wireless setup and configuration of your new DIY smart speaker or camera — though you can still use the Pi as a standard Linux computer with a monitor, keyboard, and mouse if that’s more your style. The Vision Kit also includes the Pi Camera Module V2.
Google is working in a lot of arenas to help make machine learning more accessible. In addition to the pre-trained models you can download for your Vision Kit from the AIY Projects site, Google just launched TensorFlow.js, which is a browser-based version of TensorFlow, Google’s industry-leading machine learning framework. So conceivably, now you can train your own cat-recognizer in a web browser and then run it on your self-built smart camera without having a PhD in machine vision.