Updated: We’re on the front page of hackaday! This was a joint effort among the members of the FAMiLab hackerspace. Tetsu and myself have been working on this script for awhile and just recently got the cron job running. We’ll update this post later with additional information. Thanks!
Below is a quick note from Tetsu:
We want to give people an open view into our hackerspace, but when we tried a 24/7 webcam with ustream, it was just invasive and creepy. We aren’t there to put on a show for people. We want them to come out and do stuff with us, but people won’t come out if they don’t know someone’s already at the lab.
As a compromise, we provide people a real view into the lab so they can see if people are there, but it replaces their faces with the cartoon face of Ghost in the Shell’s Laughing Man for anonymity.
Yes, it doesn’t catch every face. Yes, it catches faces that aren’t there. Yes, it’s never going to anonymize clothing, which may be more identifying than the faces. But, the way I see it, at least we’ve established a precedent at the lab that we have to face increased surveillance and increased pressure on normal people to install surveillance systems, so we may need to take extra efforts to establish norms of anonymity while staying up-to-date in tech.
Also, if any of you Hack-A-Day readers want to come by the FAMiLab sometime and are concerned about anonymity, I will personally alter your faces out of every photo before upload!
Over at FAMiLab we have on loan a Sony IPELA SNC-RX550N networked camera. It is a very nice piece of hardware; supports pan, tilt, and zoom via a Java applet and web interface. We’re often wondering if there’s anyone at the lab already, so I rigged up a little proof of concept that takes a snapshot from the camera and runs an OpenCV algorithm against it to count the faces in the photo. It worked pretty well, despite a few false positives, so I went ahead and finished it.
The code is up on my personal GitHub account, and you can snag a copy and play with it. I’m using the Python extensions for OpenCV along with a couple libraries to do the Twitter and Imgur integration. For Twitter I’m just using the python-twitter library; for Imgur I actually forked the code that Devon Meunier wrote for pyimgur and fixed a couple bugs.
Check it out, and let me know what you think. It’s very rough code, so bear with me.