AI and Machine learning concepts
are growing at an incredible rate. People who are interested in these developments need a way to practise and implement these technologies effectively. Fortunately, Google has made it much easier for us to dive into the world of AI. With the releases of their AIY kits, we are now able to create our own intelligent speaker and camera.
Google AIY kits
AIY is a combination of DIY (Do it yourself) and AI (artificial intelligence) created by Google and Raspberry. Putting the words together tells you exactly what these kits are about: do-it-yourself artificial intelligence. Google AIY currently offers two different kits.
The Vision Kit: An intelligent camera that can see and recognize objects.
The Voice Kit: An intelligent speaker with voice-recognition.
AI meets DIY
We, the researchers and builders at Handpicked Labs, bring new technologies closer and more tangible. A couple of weeks ago we ordered several Vision Kits (priced at €79) from an American webshop to discover their possibilities. These kits include the pieces for an unassembled smart camera driven by a Raspberry Pi Zero.
It took us about an hour to complete the 100 steps needed to put the Vision Kit together, but there it was! An intelligent camera capable of image recognition using neural networks. It detects everything from faces, animals, plants and objects to something as specific as facial emotions! All of this happens local, on the device itself, without using the cloud.
We thought it would be nice to make an object that detects the amount of people in front of the camera, registers their emotions and presents this information. Our latest creation, the 'Joy-O-Meter' uses the smart camera from the Google AIY Vision Kit to create a device which not only registers the amount of people in front of the camera, but also measures their happiness.
Google’s kits are a perfect way to learn and experiment with AI on a small scale without necessarily spending a lot of money. Google makes AI accessible, just like they made Virtual Reality accessible using the Cardboard a few years ago. Users are able to come up with their own applications and can even train their models with Google's TensorFlow machine learning software to make them even more effective. A nice way to gain experience and a lot of fun!