public
Description: Like walking, but with paper
Homepage: http://walking-papers.org
Clone URL: git://github.com/migurski/paperwalking.git
name age message
file README Wed Jun 03 23:17:04 -0700 2009 A few more words... [Michal Migurski]
directory decoder/ Wed Nov 18 22:36:09 -0800 2009 No longer messing with the QR code, maybe this ... [Michal Migurski]
directory site/ Sun Nov 15 23:17:02 -0800 2009 Added endpoint description comments for the oth... [Michal Migurski]
README
Paperwalking is the codebase behind http://walking-papers.org.

There are two main components here.

1. Site

The user-facing website is written in PHP and MySQL, and has been developed
to work adequately well in a commodity shared hosting environment, such as
the Pair.com "webmaster" account I use to host Walking Papers.

2. Decoder

The ugly math bits are done in an offline process that consumes a queue
of freshly-scanned images from the main site, runs them through the
image-recognition algorithm, and posts back georectified image tiles
for editing. You can run a bunch of these in parallel to make jobs go
faster, and they should be percetly fine on small EC2 instances or
a box plugged into plain old residential DSL.

It's worth mentioning that the image recognition part of the work relies
on a NON-FREE, PATENTED ALGORITHM called SIFT:

    http://en.wikipedia.org/wiki/Scale-invariant_feature_transform