Re:Lixil AI-Based Toilet Analyzes Shape and Size of Feces wi...
by awinter-py on Wednesday October 30, 2019 @04:09PM
(#21402007)
80% accuracy is potentially fine for a time series, assuming the errors are random rather than specific to a user. You can denoise this stuff, and the recognizer should get better as the sample size grows.
So why haven't these kinds of toilets taken off? Too hard to modify existing toilets? People don't care? Technology isn't there?
Seems like for those that are really interested, it isn't _too_ hard to just manually keep a log (pun intended).