Apps: Investigate testing model performance #700
Description
It's pretty clear from the graph when the inference is happening every 250ms. Can we improve performance? Is it fundamentally the ML performance that's at issue or is it the page re-rendering that's causing the problem? Let's minimise React re-rendering as a first step (e.g. move more to the store with narrow subscriptions).
If it's the ML, is there a mobile-optimised version we could use?