I can't help but think of a talk from pycon 2012:

Militarizing Your Backyard with Python: Computer Vision and the Squirrel

He used some python bindings for computer vision libraries to train his
cameras to recognize a squirrel. Then he shot the squirrels with water when
they tried to eat the goodies in his garden.

Might help, might not. Fun watching either way.

On Thu, Mar 7, 2013 at 11:59 AM, Kyle Banerjee <[log in to unmask]>wrote:

> I know about Google goggles, but I need something I can train using lesser
> known features (i.e. local stuff that's important to us but that others
> don't really know about) to help with processing of archival photographs.
> Does such software exist in a form that would be accessible to libraries?
> Thanks,
> kyle