It's really easy to forget people's names. ( https://www.youtube.com/watch?v=DYHyZHzRTFo )
Plus, being able to scan someone's face and get personal info from them from a centralized database is really cyberpunk.
What it does
NSAR shows names and other details next to the faces of people you look at with a HoloLens.
How we built it
The HoloLens detects faces and sends the images to a server that processes them and uploads the image and checks recognized faces on facebook. If a face matches a facebook account the user is friends with, details are send from the server to the HoloLens, which displays them next to their faces. Details are scraped from Facebook. The Facebook graph API was heavily used.
Challenges we ran into
We originally rolled our own machine learning with data collected from facebook profile pics, but it turns out it's hard to accurately detect faces with only one sample per person.
Programming in C++, UWP, and DirectX for a hackathon is hard, especially when unfamiliar with the Windows platform.
Accomplishments that we're proud of
We got it working! And it's impressive how well the face recognition works.
What we learned
Facebook somewhat cares about privacy.
What's next for NSAR
Faster face recognition and some dystopian government to sell it to.