Oct.4, 2013

Wish you could sit at home and visit all the places in the world? Google has been working hard on creating satellite and street maps to let you take a virtual trip around the world. And now Bryan Russell at Intel Labs and colleagues at the University of Washington in Seattle have developed a computer program that could automatically create annotated 3D reconstructions of tourist sites. All it needs is Wikipedia and online photos.

The approach is completely automated. First it downloads a set of images from Flickr by querying for the site name and then automatically reconstructs a 3D model using the freely available VisualSFM package, followed by PMVS to generate a dense 3D point cloud.

Then the program crawls through Wikipedia and to find nouns that names objects and link it to the reconstructed 3D geometry. Finally the relevant photos are then added to the appropriate locations on the model. So when the user scrolls through the Wikipedia, it will fly you through to the corresponding location in the scene. Similarly, when you click on an object in the visualization, it highlights the corresponding descriptive text on the Wikipedia page.

This approach works very well with the well-referenced popular site such as the Sistine Chapel as it requires lots of images and good text, but do worse for other sites that are not as well documented. Popular landmarks have often more pictures and description online, so the system retrieves better result with these texts. "With growth in photo and text corpi, the system will work "as is" for more scenes as the underlying data improves." says Russell. The team's goal is to improve 3D reconstruction algorithms and text parsers and further improve applicability of 3D Wikipedia.


Via: NewScientist


Posted in 3D Software



Maybe you also like:


Leave a comment:

Your Name:


Subscribe us to

3ders.org Feeds 3ders.org twitter 3ders.org facebook   

About 3Ders.org

3Ders.org provides the latest news about 3D printing technology and 3D printers. We are now seven years old and have around 1.5 million unique visitors per month.

News Archive