[TriLUG] this is a serious question - how to detect things in photos with Linux
Max TenEyck Woodbury
max at mtew.isa-geek.net
Sat Dec 8 09:22:00 EST 2012
On 12/07/2012 10:55 PM, Greg Brown wrote:
> Background: I have been tasked with building the technical end
>of a 21st century photo booth, the kind you sit in that used to
>spit out four photos, for a company on the Outer Banks. Instead
>of printing out photos said company wants to upload photos to
>their Facebook page, website, etc.. perhaps even Tweet them.
> Easy enough, right? (famous last words) Get a Linux
>single-board-computer (one with wi-fi to connect to the office
>network and Internet), a webcam, write some simple scripts and
> But that said, being as this will be an automated process and
>people are people I need to come up with a way to detect boobs,
>butts and junk (yes, I'm serious) to filter and objectionable
>pictures before they are uploaded.
> Has anyone done anything like this? I've seen some projects
>on-line that shoots water at squirrels eating birdseed and
>whatnot and maybe that's the way.
> Does anyone have any ideas?
As mentioned in other threads, a UPF is a necessary part of such a
system, but another filter could useful for pre-screening. In
particular, A face recognition procedure might be useful to sort the
stream into known clients, who can presumably be trusted with no or
minimal UPF review and unknowns that need closer inspection.
More information about the TriLUG