Skip to content

Comment on Machine Learning on the new micro:bit

Comments

FYI, this runs a realtime custom keyword spotting model in 31K of RAM and 47K of ROM (including all DSP code and Neural Network) on the updated micro:bit three times a second. We can do a lot these days on embedded devices!

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.