Friendly

N|Solid

Build Status

Friendly overlays features from all your friends to create an image of what an average person in your friend group looks like.

Made for HackRU 2019.

Installation

Friendly requires OpenCV v3 to run.

Install the dependencies and start the program.

$ mkdir friendly
$ cd friendly
$ git clone
$ pip install -r requirements.txt
$ python run.py

Benefits of the Project

There are limited size datasets available for research purposes that are properly tagged with personal data such as a user's age, gender, location as well as other metadata. This app will be released on Facebook as a game users could play to find out what their average friends look like, and the image data would be sent along with metadata tags would be collected in a database for use in research.

Parts to the Project

  • Selenium Web Scraper
  • Facebook Graphs API
  • OpenCV + CNN Face Overlayer

Selenium Web Scraper

  • Facebook Graphs API does not allow the retrieval of user friend lists unless the app undergos review by Facebook. This workaround asks a user to enter their account and navigate to their Facebook Friends List so the app can retrieve user's friend's IDs.
    • The web scraper uses chrome binaries to open a webpage where the user navigates to list of friends
    • This list of friends is downloaded and stored

Facebook Graphs API

  • Facebook profile photos are public domain so can be accessed by anyone as per Facebook Policy
  • The Graphs API interacts with a developer only API to download facebook profile images using a user's ID

OpenCV + CNN Face Overlayer

  • Using OpenCV, I use builtin libraries to estimate the user's facial pose
  • The user's facial pose is corrected and overlayed with images of all other friends

Built With

Share this project:

Updates