Deepface - facebook face recorgnition

Hsg

who dares wins
Local time
3:59 PM
Joined
Dec 28, 2013
Messages
641
Deepface is a face recognition software with 97% accuracy that is already or soon going to be used by facebook to tag photos. if your face appears in some uploaded photo, by anyone, you get an e-mail notice and then you have the option to blur your face in the photos or let it be.

If you share your "street photography" on facebook, your subjects are going to get recognized, and then they'll get an email notice, telling them that they appear in photos uploaded by so and so user, so if they didn't like their photos in your 'street shots', they have the option to blur their face in your shots.

And, its not going to be long for this system to be part of almost all image hosting sites, because its natural for people to be curious on who's photos they appear.

Goods times!

https://nakedsecurity.sophos.com/20...cognition-technology-has-human-like-accuracy/

The social network plans to use the system [Deepface] to identify its users in new photos as they are uploaded. If your visage appears in one of the 400 million pictures added to the network each day you'll receive an email from Facebook alerting you.
If you are not happy about appearing in your friends' timelines - and it is only your friends that will see it - you'll have the option to blur your face and retain your privacy.
 
I bet this backfires on FB if they implement it.

Living here in tech savy Seattle. Folks are (mostly) hyper vigilant when it comes to controlling their online presence.
FB is already sliding in their position with artists, bloggers, and "social broadcasters".
I bet they won't actually implement this in the way discribed in the OP.
People will drop FB profiles to be hidden rather than respond to non stop emails.

I dropped FB January 26th 2013 after being an early adopter.
Since then nearly all of my local friends have dropped their personal profiles.
Some have developed business pages.
 
If the face-scanning becomes commonplace across internet hosting services, etc., and it gets to the point where photographers are being impeded from artistic expression, there will likely be a business case for an internet service provider/host to offer hosting that will not be face-scanned/tagged by them. Or at least won't allow the subject to alter the image.

I can see this face-scanning become a big deal for any content aggregator, as it will allow discovering social relationships and developing a stronger profile about who you are - information to be aggregated and sold to interested parties (advertisers). Maybe FB will blur your likeness, but it will already know that it's you in the photo, as well as anyone else present. Combined with GPS metadata tags embedded by most phone cameras, they'll also know where you were photographed, and when...

Back to giving subjects the ability to blur their likenesses in other people's photos: it's an interesting situation. Of course it depends on the country, but if we use the USA as an example: photographing someone on the street is not a privacy violation. Things get a bit more complicated once you put that photo online, particularly if you're using it to promote a product or service, such as your for-hire photography. But there likely will be a degree of 'fair use' leeway in finding a common ground between artistic expression and accusations of privacy violations.
 
This sounds like bad news .
I must admit to always blurring registration plates in any cars which may appear in my shots.
I understand that this information can be used for illegal activity.
 
yet another reason to get out.

too bad FB is also becoming more common place for example to sell stuff (buy and sell groups), or like my employer, share information to employees that can be important (private company group).
this social media realm is spreading to new venues of life, its not just pizza or holiday pictures anymore. will be harder to stay away even if wanting :(
 
I'll have to keep an eye on this. We mostly use FB (a single account for Wife+Self) to keep in touch with the children and their families, not least those in Brisbane.
...
I must admit to always blurring registration plates in any cars which may appear in my shots.
I understand that this information can be used for illegal activity.
Likewise.
 
Facebook couldn't have developed this software without using its database of user photos. So facebook values your photos because its raw data to be used and abused.

And its not hard to imagine that everyone, from spy agencies to police would want such a software to use and they will pay a lot of money for it. facebook says it won't sell this software to police and spy agencies but that sounds rather weak.

In short term not much will change with this software at work, but as its adopted by others and its news spread, it will make people weary of who photographs them, that will have an impact on candid street photography.

In long term, well, in long term no one knows what will happen, but one thing is for sure, privacy is a thing of the past.
 
Isn't this software simply akin to that already in use by casinos and some shopping mall companies? At least with FB you can opt out of the whole thing and while I can simply not visit a casino, shopping malls are unfortunately fast becoming the only option among an ever narrowing field of options when you need to go shopping.
 
The casino face detection is manually coded into the system for each individual.

With Deepface, the software itself recognizes the face by profile and then detects it in other uploads. Its on a completely different scale. Deepface is artificial intelligence that does not need human input, it works by itself going through the photo uploads and then matching them with existing facebook database.

Artificial intelligence can now "see" and "recognize".
 
Hsg, Facebook gets data for 'free' (images and names associated), but the facial recognition image processing algorithms have been around for a long time. There are commercial libraries for it, years of published research papers, etc.

The recent Deep* AI talk is really describing functionality which has been around for a while. There are certainly advances in pattern matching and efficiency wins but little if this is really 'new'.
 
Back
Top Bottom