Comments on: Info on Wired Mention of Stanford Project Geocoding Photos?

Question: Info on Wired Mention of Stanford Project Geocoding Photos?

Jeff Howard — Tue, 30 Nov 2004 08:31:30 -0800

Last month in Wired, David Weinberger mentioned a research project at Stanford that "compares photos with shots of known locations, automatically annotating snaps with information about where they were taken."

Anyone know what he's talking about? I can't seem to turn up any more information about the project.

By: Smart Dalek

Smart Dalek — Tue, 30 Nov 2004 09:04:05 -0800

Weinberger's speculating on equipping the camera some form of geographic reference; while he doesn't say how it would be done, he envisions a system where you could take a photo at any location and the camera could plot the coordinates and note the spot where the image was taken.

Let's say you had photographed a stray dog sniffing a fire hydrant with a digital camera, and you later posted the image on the web. The camera would already have a record of which part of town (or whatever city or suburb) you had taken the photo - most likely through a network connection to a GPS databank and cartography archive, though it's not specified by Weinberger. We'll just say the camera is...well..somehow aware of its surroundings during the moment you open the shutter. Because of this, it can record the information onto the digital image, either as a text record ("XYZ, corner of 5th & Main"), or, with advanced software or scripting, the image could be posted online and it would automatically plot itself on a map as to where it had been taken. Overall, it's a fascinating idea, but while it's not entirely impossible, there are a tremondous number of ponies that are required to make this work.

By: smackfu

smackfu — Tue, 30 Nov 2004 09:07:09 -0800

It might be this guy. It depends on having the GPS coords of the photo, and then using that to "automatically annotate" the photo with a place name (if I read the abstracts correctly). The tricky bit is organizing the photos into groups/events and giving names that match how you would name it.

By: Jeff Howard

Jeff Howard — Tue, 30 Nov 2004 10:33:48 -0800

I was envisioning a system that could analyze a photo without benefit of GPS, recognize that the Statue of Liberty is in it, and then find other GPS-tagged photos of the Statue of Liberty, thereby inferring that your photo was taken on Liberty Island.

By: smackfu

smackfu — Tue, 30 Nov 2004 10:49:18 -0800

Yeah, me too. But on a reread it seems clear that's not what the article is talking about. The Stanford research is only mentioned in the context of what could happen upon adding GPS to future cameras.