Raspberry Pi Foundation Launches Scratch-Powered Machine Learning Projects
Designed to ease a new generation into machine learning, the projects are created in partnership with Machine Learning for Kids.
The not-for-profit Raspberry Pi Foundation has announced a partnership with Machine Learning for Kids' Dale Lane to introduce a new generation to deep-learning and artificial intelligence using the approachable Scratch blocks-based visual programming language.
"Machine learning is everywhere. It’s used for image and voice recognition, predictions, and even those pesky adverts that always seem to know what you’re thinking about," writes Daragh Broderick, content assistant at Raspberry Pi subsidiary CoderDojo. "If you’ve ever wanted to know more about machine learning, or if you want to help you learners get started with machine learning, then our new free projects are for you!"
While the majority of machine learning projects are aimed at a more mature audience and text-based programming languages, the Raspberry Pi Foundation has enlisted the assistance of Dale Lane to adapt projects from the Machine Learning for Kids website for inclusion in its official projects portal. Each project is designed Scratch, the block-based visual programming language developed by MIT Media Lab's Lifelong Kindergarten group in 2003 and designed to break programs down into jigsaw-like pieces which fit together in a logical manner via a physical, rather than textual, syntax.
So far, the Foundation has launched five projects: Smart Classroom Assistant, which uses natural language processing to recognise commands and activate or deactivate an on-screen fan and light; Journey to School, which uses survey data to train a machine learning model to predict students' routes to and from school; Alien Language, which turns sounds into commands for an on-screen character; and Did You Like It, which introduces the concept of sentiment analysis.
"When we hosted Scratch Conference Europe this summer, machine learning was the talk of the town: all of the machine learning talks and workshops were full with educators eager to learn more and find out how to teach machine learning," explains Broderick. "So this is the perfect time to bring some free machine learning resources to our projects site!"
While the choice of programming language and overall complexity means the projects are better-suited to absolute beginners, they represent a fantastic way to introduce the concepts to a new generation. Once the core concepts are well-grasped, students can move on to deeper projects such as reinforcement learning for self-balancing motorbikes, hand-gesture classification on the Google AIY Vision kit, and move to alternative platforms including the Nvidia Jetson Nano, Xilinx DNN Developer Kit, and Erle-Brain 2.
Links to the projects are now available on the Raspberry Pi nlog; additional projects can be found on the Machine Learning for Kids website.