insight3d / Image based 3d modeling

I been playing with photogrammetry solutions a few years back, but it never stick with me. Perhaps it was too much work for less of a demand in my case, but you got to check out this new opensource project by Lukas Mach named insight3d.

What is cool about it is it is opensource and free… this is always nice, and it seems to work rather well from what I gather and the tutorial posted by Lukas. If you do check it out let us know what you think. I’ll try to use it on a current studio project and share the results with you.

Villa in Prague reconstructed using insight3d

insight3d lets you create 3D models from photographs. You give it a series of photos of a real scene (e.g., of a building), it automatically matches them and then calculates positions in space from which each photo has been taken (plus camera’s optical parameters) along with a 3D pointcloud of the scene. You can then use insight3d’s modeling tools to create textured polygonal model.

7 replies
  1. Samuel Conlogue
    Samuel Conlogue says:

    That’s great! I was just writing a blog post about this very technique and I missed insight3D. Great find! I Particularly like that it is Open Source! There are new tools coming for Max that allow you to import and use cloud data native within Max and Mental Ray so I imagine this being very useful.

  2. Willem
    Willem says:

    And now with the Kinect making it into this realm also, we’re seeing the beginning of exciting times ahead.

  3. Tobias
    Tobias says:

    Have you already done anything with insight3d?

    I wasn’t able to compile it on Linux and the author doesn’t react on eMails. 🙁

  4. Avanindra
    Avanindra says:

    It’s similar to Bundler, except that it does assume any information about focal length of images. Though it’s not as robust as bundler, it fails in case of planar surface points.

    Also it does not create dense surface, it only outputs sparse point sets.


    You need to redefine CV_ERROR macro (remove the goto exit; statement) and define labels _BEGIN_ and _END_ in cv_extensions.h and the code would compile without any error.

    Also clean all sub-projects like sba, ann etc. then rebuild.

    at some places you would need to explicitly give the default parameters.

  5. Avanindra
    Avanindra says:

    One correction to my above comment.

    “it does NOT assume any information about focal length of images.”

Comments are closed.