Topic: 3D textured models from the Syntesized data

Report Abuse Report Abuse
PeterVrenken (Over 1 year ago)

would it be possible to generate a textured 3D model of the gathered PhotoSynth information? It would be really cool if PhotoSynt had this feature.

Even more great would be a permanent connection with Panoramio and other Geolocated picture stores. This would allow for 3D synthesizing of the whole world.

Second life be gone! :-)

Greetings from the sunny Netherlands,

Peter Vrenken
Hi Peter -- I think there is a fair bit of active university research into generating textured models from free-form photography. It's not something we're working on directly right now.

Re 3D synthesizing all the interesting parts of the world from the photos on the major photo sites: This is something that often comes up, in part because it was a guiding vision for the original Photos Tourism work from the University of Washington, and was also discussed at TED by our architect, Blaise Aguera y Arcas. Since those days we're discovered that all the world's photos don't actually include much variety. People don't shoot the boring bits, unless they are purposefully documenting the space. For example: We made the best synth we could from all photos on Flikr matching "Notre Dame". It wasn't nearly as complete as what one experienced person could do who knew they were shooting for Photosynth:
Bellerophon2200 (Over 1 year ago)
Hello Peter,
   I'm trying to get a point cloud to go into fallout 3 through all the programs; exporter>meshlab>blender>nifscope>GECK>Game Level.  Fallout 3 already has a physics engine (gamebryo) and realistic user editable avatars.  However, I'm a far from actually getting it into the game.
michaeldenis (Over 1 year ago)
Cool idea Bellerophon2200, I'd be very interested to learn more about your project, especially as you become more successful with it!

To comment on what David said almost a year ago, I am unsurprised that the types of photographs available from general photography (whether tourist or commmercial) lack the precise features that make a "good" synth and a "good" point-cloud.  My sense is that the demands of Photosynth and pointcloud-based photography in general is a frontier.  We have an idea of what should exist out there, but we still lack the tools and experiences to draw a solid map of what lies on the other side.. interesting posts.
Nathanael (Over 1 year ago)
Michael, my interpretation of what David was saying of crowd based photo collections not being so useful for 3D reconstruction was that this was the case because people only shoot landmarks from a very limited range of positions, never bothering to walk all the way around a building and shoot photos of the sides or back.

In that sense, it isn't so much the case that the photos in existing online collections don't have the right features, but rather that they don't come from a deep enough range of positions. I think from the front of the building or monument, such collections probably have all the features that one could wish for, in order to construct the front facade (although not perhaps as thorough as systematic coverage designed to build fine grain pointclouds of every light fixture or relief on the front of said building).

The question now is only how to fit all the photos into one synth.

Apologies if that is what you intended to communicate to begin with.