Wyze should look into this cutting edge solution.
I read this article recently about a new way to do AI learning that takes hardly any data input:
Doing something this way could potentially allow PD and any other detection for that matter to be done without huge computational dataset foundations. Instead of continually basing everything off thousands or millions of sample videos, the AI can be trained in this manner with very few examples.
I’m not saying it is something EASY or whatever right now, but they’ve demonstrated that it is possible to have AI detection work effectively (locally or on the cloud) without needing tons of data and computational resources. Wyze recently bragged in their Headphone video that they have a bunch of super smart PhD employees. Maybe someone can understand this enough to make use of it and save Wyze from having such huge expenses.