alt Hacker NewsI wish something like this existed that was completely offline. I'm face blind (prosopagnosia) so being able to feed an offline database photos of friends so it can recognise them would be great.
Accessibility shouldn't require giving up privacy.
Replies
I find myself asking this every time a new software product is released. "Nice, but why can't this be usable without an account and a tether to the developer's cloud?"
It’s possible to make it offline but one would still need a phone for the object recognition software…that software would have to be what stays offline. Definitely doable.
Meta has to release the hardware, then we can get started on alternate OSes.
It's a shame open source hardware isn't a thing in this area, but we've been here before. (Buying locked down devices and installing alt OSes.)
Probably quite feasible. Immich on my little raspberry pi is able to do facial recognition for 50k images over night.
Honestly, if any big tech would implement it this way, it's likely Apple. Their image face recognition in Photos currently is fully on device from what I understand and it is set by who you associate it with locally.
Sorry that you have to deal with this condition. What method do you use currently to help with recognising them?
What ever model digiKam is using for face detection seems to be very accurate in my experience. It is open source and works fully offline.
Apple is the only company doing this sort of processing
This shouldn't be hard to do, facial recognition is pretty easy to run locally.
Literally my thoughts. That would be so helpful, but I will not give any data-hungy company access to people around me.
I recently heard the best way to explain faceblindness: Apples.
Can you tell apples apart? Yes, sure, if you put two apples next to each other, they look similar, but there are differences.
But could you recognize that specific apple among 50 similar ones?
If an apple addressed you on the street, could you remember where you've seen it before?
That is how it feels to be faceblind.
There are workarounds, but they are context-dependent and error-prone.
That apple with red hair and a beard? Sure, that's the colleague from the office next door. But was that the same apple that waved to you in the city yesterday?
The only green apple among red ones? Easy to recognize. But only after some awkward misunderstandings you realize that there are two of them.
And changes of hairstyle are a real problem. I once wondered who that new colleague was during lunchbreak. I was about to ask her, when she said something (unrelated) and I recognized her voice. I had worked with her for 10 years, she had colored her hair.