PhotoSketch Released – Make 3D Buildings from Photos

Back at the end of April, I mentioned the impending release of a plugin for Google SketchUp that would allow you to quickly create 3D models from photos. Yesterday I received an E-mail from Brainstorm LLC, the makers of PhotoSketch, that their plugin is now available (Windows only for now, Mac version in the works). The plugin is available as a 30-day free trial (see pricing). PhotoSketch works by letting you use a succession of overlapping photos of buildings and, using PhotoSketch and SketchUp, you can easily use the photos to derive 3D models from them. Watch this video tutorial to see how it is done:


Now, where would you get a succession of photos in a drive-by or fly-by situation I wonder? Could you perhaps use photos from Google Street View to create 3D models? Or, how about YouTube videos in a drive-by scenario? It will be interesting to see if modelers, who are avid 3D building creators at the 3D Warehouse, latch on to PhotoSketch to create buildings faster. [UPDATE: Brainstorm tells me that Street View imagery is "generally not sharp enough, nor does it necessarily have the amount of overlap and coverage *around* the building to make it useful. Another problem is that we cannot calibrate their camera because we don't have access to it. Therefore, it is not possible to correct for radial lens distortion." So, apparently you need more accurate photo data to use PhotoSketch properly.]

About Frank Taylor

Frank Taylor started the Google Earth Blog in July, 2005 shortly after Google Earth was released. He worked in 3D graphics for many years and was very impressed with this exciting product. Frank left in 2009 to circumnavigate the earth by sailboat as part of the Tahina Expedition.



Comments

  1. That’s pretty wild! There’s a cathedral near here, it’s a bit ambititous but I might have a go at making a model of that.

  2. dinther says:

    I watched the video and although I am a techie but I am not even going to try this. This process looks extremely complicated and in the end you have to draw the buildings yourself anyway.
    These people seem to be more onto it
    http://www.photo-to-3d.com
    http://www.cs.uiuc.edu/homes/dhoiem/projects/popup/
    http://homes.esat.kuleuven.be/~konijn/3d/
    Many years ago I saw a great application that used multiple photos of the same subject and turned it into a good 3D model. Only a little user interaction was required. If only I could remember the name.

  3. Maybe the sites like http://norc.pl would be sharp enough…. :)

  4. In response to dinter’s comment about having to draw the buildings yourself anyway, realize that all users of standard modeling tools (such as SketchUp) have to build models manually. PhotoSketch simplifies this process by adding a layer of images to guide the user.
    The “automatic” techniques cited in the first and third links in dinter’s post are nothing more than traditional dense stereo techniques. The output is a continuous mesh of 3D points on the entire surface. It is like throwing a net over the objects. You basically have a “take-it-or-leave” approach to modeling because this data is difficult to segment or edit to generate useful models. Also, the output of dense stereo generates models with too many polygons that degrade performance on web-based applications such as Google Earth. Finally, the second link is nothing more than eye candy that does not create an actual volumetric model from a single image. Instead, it mimics a 3D model by cutting and folding an image to give the “illusion” that you are flying into the image in 3D. No serious modeler would consider this a solution to replace conventional modeling tools.

  5. the application mentioned above maybe Canoma. it created 3d models from 2 photos or more.

Leave a Reply