OSM 3D in a CAVE


Today we made it: viewing OSM 3D data in a CAVE!

-grab OSM data
-use OSM2WORLD to convert OSM to .obj format
-convert .obj with DeepExploration to .wrl (the exported .obj has some issues, even meshlab cannot load it)
-use our DAVE with OpenSG to view the .wrl :slight_smile:
DAVE: http://www.cgv.tugraz.at/DAVE

to bad we had no camera with us, next time we will.

There are still a lot of issues, which are only seen in a virtual 3D world of real size, e.g. trees far to big, tunnels for ways under sidewalks, buildings without windows,…
And some issues with transparency and colours.
But bus stop signs and street lamps do look very nice, also barriers like fence and hedges.

One small, important information: although the process is known now, this is NOT streamed. We still need to fetch .osm manual and convert with 2 tools until we can load and view it. We cannot “fetch data from osm server and view on the fly if we move forward in the model”.
Also some limits of area size does appear, as converting takes memory and time.

AND last: this is no official project, it is just for fun on my spare time. Luckily I am able to use the DAVE.


Very cool. Please do post a video.

What you guys need inside your cave is an omni-directional treadmill. The best version I have seen is this Swedish design: http://www.mseab.se/The-Virtual-Theatre.htm

On a related note, I think a ground-level browser of OSM 3D data where you can walk around like in a game would encourage people to explore areas and edit them based on their direct (ground-level) experience. This is possible with the OSM2World application but that was build for debugging rather than as an end-user experience. The same limitation of pre-loading the data also applies.

Wow, that’s really cool. Like Oliver, I would eagerly watch a video recording. :slight_smile:

And of course, if you have more suggestions for improving OSM2World based on your experiment (e.g. regarding the issues you mentioned), feel free to contact me with details.


Ok, I send a mail to peda already yesterday.
As with a redo of model generation:

  • window config enabled, but not in model visible/exported at all
  • tree size and model at all
  • tree min_height not supported (trees e.g. on buildings)
  • landuse (different colours for the ground in cities)
  • Traffic lights?
  • traffic lamps are available, but do not support height
  • far to less 3D data in OSM :wink:

Later on I will tryout the video.



Video made:

That is a extract from the GRAZ Area of OpenStreetMap (http://osm.org/go/0Iz_~8SX--), converted to .obj with OSM2World, converted from .obj to .wrl with DeepExploration, spiffed up a bit with OpenSG shaders (textures lightning and shadows), and viewed in our DAVE (http://www.cgv.tugraz.at/DAVE).
MAde with a typical FullHD video camera, without tripod.
Video has 130 MB, is shaky, dark, double images, it does NOT show at all what the person with the shutter glasses do see!
But it shows in rough the capacity of a full immersive virtual environement with good models, and lack of some features of the OSM model.

E.g. look at the trees, the entrances should be in 2m height!

So far we have a process defined for our tests to be able to extract any region of the world out of OSM (if not to big) and get into the DAVE to be viewed in 3D.
But as it is just a hobby project, it is not a project with funding or anything else!

This was made real with the support of TUGraz, CGV and Fraunhofer Austria, Visual Computing