Data Collection
run.py handles all of the RESTful/outward-facing requests (eg. /verify/<username>).

collect.py and verify.py have medium-specific methods for verifying that a participant has followed us, and collecting their data.

• /collect is run either manually or as a cron job every 15 minutes (when turned on)
• it looks for any usernames with collected=0 in the usernames table

extract.py feeds subroutines to the /collect regime for feature extraction.

util.py has various database and other helper functions.

Analysis

• eda-instagram and eda-twitter are the analytical frames except for Bayesian stuff
• *-bayes.R for Bayesian logistic regression and convergence checks
• bgfunc contains all the heavy lifting for analysis and processing
• face-detection compiles, trains, and verifies face detection algo
• mixture-models has code for Kalman filter, GMM, and HMM