Perfect, I'm looking forward to a world where everyone has their name floating above their heads like in World of Warcraft. Saves you having to remember stuff.
This detects a face and displays the attributes it calculated about it. That's pretty interesting though from the title I thought it was something that looked up people and displayed their name, etc.
Pretty cool though ultimately it just takes the Halo Lens input, sends it to the Microsoft Face API[1] and displays its results.
It does try to identify names, actually. Looking at the source[1], once it has used the Oxford API to find where faces are in the frame (and their approx age and gender, etc), each individual face is then also queried against the MS emotion API and an OpenFace[2] instance. Given the source, it looks like that instance is meant to be trained on staff members and report their organizational hierarchy.
What would be super neat would be ripping out the OpenFace query, and replacing it with a lookup via Facebook[3], so it could actually just show the names for anyone I was FB friends with.
Yeah, name and other attribute lookup is feasible (and seems inevitable). Also, this prototype only claims to take a snapshot and submit it. Needs to perform the check in realtime against video input.
It is not my call, I defer entirely to The Machine. We can not understand its reasoning, only attempt to understand how it became what it is. Serious question on this topic though, does anybody know if this facial recognition library has been completely open sourced? Apologies if the answer is glaringly obvious or has already been answered in this thread.
I'm honestly astounded that this works. The color camera on the Hololens is 1280x720, making it useless for most tasks. I love this device so much, but man, such a misstep.
The color camera on the Hololens is 1280x720, making it useless for most tasks
This seems.. somewhat extreme.
I built a (prototype) hand-tracking system using a 800x600 webcam[1]. In Javascript. Using a Flash->Javascript bridge. In 2009. I'm surprised that one can't find useful things to do with something that appears somewhat more capable.
Well, "most tasks" here is really just the things I've wanted to do -- not exactly an unbiased sample. Things like tracking objects, reading QR codes, matching more complicated hand gestures (than the built-in ones). Unfortunately, a lot of the really interesting things that could be done involve the high-res IR cameras, which aren't exposed in any way to userspace.
<sarcasm>Yes, I'm glad people panicked about Google Glass possibly allowing for this, but by the time Hololens gets to try it, it's inevitable and cannot be stopped.</sarcasm>
Google specifically disallowed apps from doing this (though you could sidelide them I guess), and lots of people complained that it was a useless gesture.