Plug deployment
Michal Migurski
I thought you might be interested in a somewhat weird deployment environment I've been experimenting with for Walking Papers:
http://mike.teczno.com/notes/walking-papers-plug.html
It's totally applicable to some of the field use, "tent network" uses we've been kicking around. The part that the plug computer does worst at is the image recognition. Otherwise, it's quite a serviceable machine!
-mike.
----------------------------------------------------------------
michal migurski- mi...@stamen.com
415.558.1610
Marc Pfister
On Mar 8, 12:04 am, Michal Migurski <m...@stamen.com> wrote:
> It's totally applicable to some of the field use, "tent network" uses we've been kicking around. The part that the plug computer does worst at is the image recognition. Otherwise, it's quite a serviceable machine!
Mike,
Is the slow part the SIFT or the QR recognition? Or both?
Marc
Michal Migurski
> On Mar 8, 12:04 am, Michal Migurski <m...@stamen.com> wrote:
>
>> It's totally applicable to some of the field use, "tent network" uses we've been kicking around. The part that the plug computer does worst at is the image recognition. Otherwise, it's quite a serviceable machine!
>
>
> Mike,
>
> Is the slow part the SIFT or the QR recognition? Or both?
Hi Marc,
It's definitely SIFT that's the slow part. QR code recognition works quickly, I imagine because it's been optimized for use on mobile devices. I'm on the lookout for something that could replace SIFT.
Marc Pfister
> It's definitely SIFT that's the slow part. QR code recognition works quickly, I imagine because it's been optimized for use on mobile devices. I'm on the lookout for something that could replace SIFT.
You've already seen it, but for the benefit of everyone else I was
getting pretty good results just using the locator points built into
the QR codes for georeferencing the image:
http://drwelby.net/qr-osm.html
http://drwelby.net/qr-osm2pt.html
Marc
Marc Pfister
tentnetting? That made me thing of something like this:
http://www.flickr.com/photos/drwelby/4417352687/
This is a bottom view - you could just set it down on top of a battery
and it would boot up and do its thing. It would be a similar size to a
car battery, and the slots would have spring-loaded contacts and would
accommodate different battery terminal spacing.
Maybe run a Gumstick in it or something?
Marc
Michal Migurski
> On Mar 8, 8:03 am, Michal Migurski <m...@stamen.com> wrote:
>
>> It's definitely SIFT that's the slow part. QR code recognition works quickly, I imagine because it's been optimized for use on mobile devices. I'm on the lookout for something that could replace SIFT.
>
> You've already seen it, but for the benefit of everyone else I was
> getting pretty good results just using the locator points built into
> the QR codes for georeferencing the image:
>
> http://drwelby.net/qr-osm.html
> http://drwelby.net/qr-osm2pt.html
I remember this! I've been using the SIFT corners as a way to determine paper size based on aspect ratio - we currently have six paper size/orientation options. I have modified my QR use slightly, and I've found that it's not always possible to find the code in a large image - some source files, it works great, others, not so well. I can't yet figure out what the difference is, because I can usually isolate just the QR part of a failing image to get it to work.