The reviews of this are rather mixed but I haven't seen any long-term reviews in which the author was patient enough to give the camera weeks of yes/no training on faces. I've been training Arlo Pro's and found they become more accurate over time in terms of recognizing people/vehicles/animals and filtering out false-positives - BUT Arlo does not do facial recognition yet. Nest is using Google's Face net. Nest Cam IQ review: Impressive but expensive
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Anybody using the Nest IQ camera with trainable facial recognition?
- Thread starter calisnow
- Start date
-
- Tags
- Autonomous Vehicles
arlorose
Member
I have the Netatmo Welcome which I loathe.
So I'm essentially going to use everything I hate about it to design my own Welcome Camera. It won't be limited to only security features though. I have a massive list of features that I'd like it to have. My plan is to base it on the Jetson TX2 platform so I can do local inference without having to upload images to the cloud until the training. All the training will be done on the cloud where the Jetson TX2 sends it the incorrectly marked images with the corrected label..
I'm still working out the details of how I'm going to deal with the UX when it comes to user training. The Netatmo products rely on the user to correct who/what it thinks something is, but the UI is cumbersome.
I want the unit itself to get better at learning for itself.
Like lets say that a misdetection happens 10ms after it correctly detects that I'm there then it really should go 'oh, crap".
It also needs to do a better job of using Bluetooth beacon technology, and location services within the app on the phone than the Netatmo does.
So I'm essentially going to use everything I hate about it to design my own Welcome Camera. It won't be limited to only security features though. I have a massive list of features that I'd like it to have. My plan is to base it on the Jetson TX2 platform so I can do local inference without having to upload images to the cloud until the training. All the training will be done on the cloud where the Jetson TX2 sends it the incorrectly marked images with the corrected label..
I'm still working out the details of how I'm going to deal with the UX when it comes to user training. The Netatmo products rely on the user to correct who/what it thinks something is, but the UI is cumbersome.
I want the unit itself to get better at learning for itself.
Like lets say that a misdetection happens 10ms after it correctly detects that I'm there then it really should go 'oh, crap".
It also needs to do a better job of using Bluetooth beacon technology, and location services within the app on the phone than the Netatmo does.
Similar threads
