Ken Birman
CIS
My computer science group has been building very cost-effective “platforms” to host machine learning on farms or close to other kinds of data sources. The whole idea is to be able to reuse off the shelf ML solutions (computer vision tools, for example), but to customize them to process data where you capture it — perhaps robots in a vineyard, or in a dairy. Often a small pipeline of MLs will be needed, but with clever design, we are finding it surprisingly easy to repurpose ML from settings like self-driving cars to play roles on the farm — roles valuable in diagnosing animal or plant health issues, managing the farm for productivity and cost efficiency, etc. All our software is open source and free of any restrictive licensing.
KEYWORDS: machine learning, sensing technology, IoT and networks, trustworthy AI, dairy/livestock production, animal health