@stoneatti, I agree on your thought on prioritization - tackle the EVs with the biggest footprint in the marketplace first. Also like what you did with TP-Link.
But your whole second half flies in the face of what is feasible, at least when doing realtime detection. It’s not “stubbornness” - the issue is that some users, like yourself, want to believe in solution that can’t possibly work technically. The proof:
- Two companies that offer a product similar to Sense’s, but with a “training mode” as you describe, have gone out of business. One of the main reasons is that “training mode” did NOT work.
- Most users are clueless when they talk about the “database” or library of devices that Sense maintains. One of the biggest issues is variation within on and off transitions (they are never exactly the same). Doing fuzzy matching is much harder than you think and relies on a large statistical database rather than a few training cycles that users seem to thing would do the trick. If you want further discussion on the technical feasibility of “training mode” read this end-to-end first.
How Sense does “Native Detection” Work ? - User Perspective