I’ve recently got a Netatmo Welcome camera which features face recognition. After a couple of days’ training per face it’s doing pretty well.
Netatmo have a published API for the camera which allows retrieval of movement data and snapshots. Live video is also supported but seems to require a lot more scrutiny of the app and approval from Netatmo before they’ll allow it, so I’m not planning to implement that any time soon.
I’ve written a plug-in which retrieves movement data every 10 minutes and creates/updates a ‘Last seen’ variable in Vera for each registered face. If anyone is interested in trying the beta and subsequently wants to suggest additional functionality just let me know via PM or here and I’ll send you the files to try.
The API also supports webhooks for sending updates but I haven’t even started thinking about that yet.
Is there a way to reduce the data time in receiving it from 10 minutes to even seconds? Or perhaps, when a face (specific person) is detected it sends to Vera?
The webhooks api might do that, but you have to go through a much more rigorous process with Netatmo to access that method due to privacy issues. I don’t currently have any plans to do that.
The plugin polls Netatmo servers once per minute. This could probably be shortened (although I seem to recall the API has polling limits), but the camera just isn’t reliable enough to be used to trigger scenes. Quite often it will not recognise a face until someone has been sitting for several minutes, and sometimes it fails to register a face altogether.
It’s a good idea in theory, but I’m not sure the Netatmo face recognition will ever be good enough for home automation use.