Format
====================
1. Profile image features from Face++ and Microsoft Oxford API 
This is in JSON format. One user per line (there is an anonymized user_id for each user). We provide profile images from  Facebook, Twitter, Instagram, About.me, LinkedIn and Flickr. For each image, there are following features:

Features from Microsoft Oxford API
------
accent_color
age
background_color
face_height
face_num
face_width
foreground_color
gender
img_height
img_width
is_bw
obj1
obj1_val 
obj2 
obj2_val 
obj3 
obj3_val 

Features from Face++
------
plus_age
plus_age_val
plus_face_num
plus_gender
plus_glasses
plus_pitch_angle 
plus_race 
plus_roll_angle 
plus_smile_val
plus_yaw_angle


2. Profile image deep learning features
Format: 
img_id | 1000 Object features | 4096 Deep learning features

(img_id = user_id + '_SN-Name')
(e.g, '123_facebook' is the deep learning feature of user 123's Facebook profile image)