Our first goal was to use the Raspberry Pi's and camera attachments that we happened to have to recognize Facebook friends in photos. Eventually, we realized it'd make more sense as a web/mobile app, and that the implications of this software are much broader.

What it does

We scrape photos that a user on Facebook is tagged in and use it to seed Watson API's facial templating feature. Then, we use that template to match faces found in any pictures through a web application.

How we built it

We used a Python script to scrape data from the Facebook Graph API and feed it to Watson's Visual Recognition API. We have a web server that can take images and interface with Watson's API, returning the most likely matches for the faces and their corresponding Facebook ID's.

Challenges we ran into

Many of the members of our group had to dip their toes into some area of development they'd never worked on before.

What we learned

Python scripting, web development.

What's next for findr

See whether it's possible, with enough computational power, to scrape more publicly available data from Facebook to build a bigger repository of templates. Also explore other uses for this facial recognition feature, e.g.spotting friends in certain locations, helping visually impaired people identify people, pull up a brief Facebook history of someone you haven't seen in forever so you don't forget their name,etc.

Share this project: