If you’ve tried the Powerhouse Museum layer in Layar in the past few days on the streets of Sydney you may have noticed some odd quirks.
Let’s say you are in Haymarket standing right here.
You open Layar and it tells you that where you are standing is the location of the following image.
Now when we were geo-tagging these images in Flickr we made a decision to locate them on the point closest to where the photographer would have stood. That seemed like a sensible enough option as it would mean that you could pan around from that point in Google Street View or similar and find a pretty close vista. This is well demonstrated in Paul Hagon’s mashup.
In the example above, if we had geotagged the subject of the image (the lighthouse) on its exact location then the Street View mashup would not function. This would be the same for many other images -the Queen Victoria Building, the Post Office, and the building in Haymarket.
However, AR applications work in the physical world and so we have another problem. If you are walking around you don’t necessarily want directions to the place where a photograph was taken, but directions to the subject of the image – especially if the camera-based heads-up-display is overlaying the image over the view of the world. This is particularly the case with historic images as the buildings have often either changed or been demolished making the point-of-view of the photographer hard to recreate. (Fortunately the Haymarket building is still there so reconstructing the view is not too difficult).
The larger the subject, the more problematic this becomes – as the photographer would stand further and further away to take the shot. Think about where a photographer might stand to photograph the Sydney Tower (or the Eiffel Tower) for example – it would be nowhere near the actual location of the subject of the photograph. Showing this on a mobile device makes far more sense if it is the subject of the photograph that is the ‘location’.
Question is, should we re-geo-locate our images? Or geo-locate both the photographer’s position and the subject’s position separately?
Either way we need to look into how people actually use these applications more – it might be that it doesn’t really matter as long as there are some obvious visual markers.