#Timesheets#
###What it is###
Over the course of the last year (June 2014-June 2015), I have been recording my face as I've been completing my timesheets at Smart Design, in hopes of getting a better understanding of my mood and behavior performing this repetitive task.
As a summary for my performance review evaluation, I spliced together all the video into a timelapse video available here
markbreneman.github.io/timesheets
and ran facial recognition on all the video, to build data files which would document my emotions, race, age and "beauty" for the year. Lastly I also took stills from the video and aligned the detected faces into aggregate imagery of what it might mean to try and capture the timesheets experience in a more abstracted nature over time.
They end up being quite scary;
###How it works###
####Attribution#### This project is built on the shoulders of giants. It uses a slightly modified automator script from Kyle McDonald to record a users face when they visit a website, and a modified Rekognition to Processing library from Daniel Schiffman. The processing code is my own(cause who else would write something so absurd), and is separated out into two files; one for the aggregate image overlays and one for the data file recording. It could definetly be optimized into one sketch.
####Workflow#### I had the scripts setup to run and record anytime I visited the SmartWorks website. With the videos recorded overtime I compiled them in Final Cut into one 5 hour long video. Then compressed them into a time lapse of 2 min. Then exported all the frames; 3600 PNG files and renamed them numerically so each was #.png . Those PNG files were then placed into the data folder of each Processing Sketch. The Processing sketches were then run using a modified Rekognition-for-Processing library that I modified in Eclipse(reference the Rekognition for Processing -mb folder in this repo.) and analysis was done using the Rekognition Face Detection API. All the PNGs were too much for github so I've removed them, but the end data files are located in the Results folder as JSON Data.
###To do it yourself###
The modified automator scripts are located in the recording scripts folder, and can be modified with automator to work on any website (change the lines - 'set targetDomain to "domaintoRecord.com"`)
and browser of your choice(change the lines - tell application "application name").
Once the script is set to record go about your normal behaviors and it will record and save to the specified folder(on the line - set folderName to (path to desktop as text) & "folderNameDesired").
With the video saved the next thing to do is register for API keys with Rekognition, and place them in the key.txt files in the sketch folders. Next you'll have to compile the Rekognition-for-Processing-mb library in Eclipse(sorry but thats life). There's good walk through here;Processing Library template. Lastly put save the video into images using Final Cut or FFMPEG etc. and place them in the Data folders. Then run the Processing Sketches and await the fun times.