Squirrel recognition with the Raspberry Pi 3

Bulb monster. © iStock.

Ofer Dekel is a researcher at Microsoft, in Redmond, Washington. Dekel has a problem: squirrels are feasting on flower bulbs and bird seed in his backyard — a problem experienced by most North-American homeowners. Dekel has developed a solution that isn’t currently within everyone’s reach: a model for the visual recognition of squirrels for the Raspberry Pi 3, where an algorithm identifies the squirrels, triggering the sprinkler system. Dekel has put together a multidisciplinary team to develop a toolbox of Machine Learning tools and software to embed artificial intelligence on small-scale systems such as the Raspberry Pi. The challenge is that today’s hardware-intensive learning algorithms are incompatible with small, limited-capacity devices; Dekel’s team is working on “compressing” deep neural networks to fit smaller systems. Early previews of the software are available for download on GitHub.

⇨ Next at Microsoft, “AI’s big leap to tiny devices opens world of possibilities.”

Low-tech solution. © iStock.