It’s supposed to improve over time, get better at recognizing your face. That’s the whole point of machine learning.
It’s supposed to improve over time, get better at recognizing your face. That’s the whole point of machine learning.
That's technically infeasible. It can only get worse.
Could you elaborate on that? My experience matches the OP. It seems to work more quickly than it did initially.That's technically infeasible. It can only get worse.
Lolwut?
Go read a wiki on “machine learning”.
Example:
If you met two identical twins & couldn't tell them apart at 1st... would you you get better at telling them apart over time, or worse?
C’mon... think it through!
Could you elaborate on that? My experience matches the OP. It seems to work more quickly than it did initially.
I’d be curious to know why it “can only get worse.”
Gotcha.I didn’t have faith that Apple could pull it off.
Kinda like your posts.That's technically infeasible. It can only get worse.
It’s working even better for me too since launch.
Even first thing in the morning, in different lighting conditions and further away.
I like the fact that if I have my hand over my mouth (thinking kind of way), it doesn’t work
It absolutely learns. I can put my hand over various parts of my face and Face ID will fail the first time. Then I put in my passcode (which tells Face ID that it’s me it’s failing to recognize) and without moving my hand away from my face, it works next time to Face ID log in.
This is interesting.Yep. In fact, it was documented in a white paper - both successful and unsuccessful (but within a specific threshold), are used to augment the existing enrolled facial data (i.e., the mathematical representation).
(Side note: working on a neat ML project, not Apple related, but very cool stuff ...)
That's technically infeasible. It can only get worse.
I was surprised this morning when FaceID recognized me despite nearly half my face being covered by the pillow and my elbow. I knew it would get better over time, but it has improved beyond my expectations.FaceID works well except when i'm in bed and my face is half covered by my pillow
This is interesting.
If unsuccessful Face ID attempts followed entry of passcode are indeed used to 'improve' detection rates, does it mean it is best not to use other family members’ phones using passcodes? Would we make the base detection rates worse?
If unsuccessful Face ID attempts followed entry of passcode are indeed used to 'improve' detection rates, does it mean it is best not to use other family members’ phones using passcodes? Would we make the base detection rates worse?
Makes senseNo, because they wouldn't be well within the threshold of a match
I think the machine learning will have a threshold of how different a face can be until it just chalks it up as a different face entirely and therefore it will disregard it for the purposes of improving 'recognition' of your face. This is how I think it will work anyway.This is interesting.
If unsuccessful Face ID attempts followed entry of passcode are indeed used to 'improve' detection rates, does it mean it is best not to use other family members’ phones using passcodes? Would we make the base detection rates worse?