Alasdair Allan
1 min readAug 23, 2019

--

I probably could, but I’m not entirely sure it’s a useful measure. I’m guessing that most embedded hardware with a camera will be continuously inferencing from a video stream looking for ‘events,’ then reacting to those events in the data stream in some manner. So I’m guessing that continuous, rather than per-inference, load is a more useful measure. Maybe. 🤔

--

--