My nest doorbell worked reasonably well for facial recognition, but since we added Nest floodlight cams and Nest indoor cams the facial recognition has become absolutely terrible. It’s affected recognition on the doorbell too! When it thinks my 7 year old daughter is my 80 year old neighbour you know there’s a problem! And I thinks everyone who wears a hat is my dad, any anyone wearing sunglasses is one particular friend. It’s completely useless. Is there a way to wipe the database to see if it works better from scratch?
Deleting mismatches is pointless. Unless you can reassign them to the correct person it will never learn! Kind of defeating the point of machine learning if you can’t train it…
I apologize for my late response and thank you for all of your patience while waiting for a reply. I definitely understand how frustrating your situation must be, and I wanted to make sure that your concern was acknowledged. Do you still need help with this issue? I know that there's been a delay in response since you posted.
I don't need help, I need Google to fix the product's flaws...
I just get told to wipe my face database, and then to tell it when it gets it wrong. I've been doing that for months and have wasted hours and hours, it doesn't learn because you can't train it. It's a fundamental flaw of the product. I have plenty of examples of well lit well framed faces it gets completely wrong, and when you delete the wrong matches, continues to get it wrong.
Exactly, you can't help us. There is nothing we can do to fix this problem. The camera doesn't recognize faces well and there is no way to tell it that this face is this person. You need to go back to the coding team and tell them to add a assign face to library person function.
I just wanted to jump in real fast and thank you for your feedback. Please feel free to submit your feedback to us. I've put instructions for feedback below.
If you need anything else, please let me know.