Home Technical Talk

Tutorial: 3d scanned rubble prop

AlecMoody
ngon master
Offline / Send Message
AlecMoody ngon master

Hi Everyone,
I put together a tutorial covering my workflow for a 3d scan based rubble prop. It is intended for people who are new to 3d scanning but familiar with game art workflows:
http://www.alecmoody.com/rubble/

Replies

  • Mant1k0re
    Offline / Send Message
    Mant1k0re polycounter lvl 8
    Nice tutorial Alec! I've tried a few solutions for photogrammetry but it's the first time I hear of Reality Capture. What are the reasons you picked it over Photoscan for instance?
  • chrisavigni
    Offline / Send Message
    chrisavigni polycounter lvl 12
    Nice one! Tks!
  • AlecMoody
    Offline / Send Message
    AlecMoody ngon master
    Reality capture is way faster than photoscan (I own both and as soon as I bought RC I stopped using photoscan). If I had to guess, I would say RC is about 10 times faster. Processing 100 35 megapixel images for this scan on high quality settings took a few hours. I wouldn't even attempt to process images that large on my machine with agisoft.
  • Anchang-Style
    Offline / Send Message
    Anchang-Style polycounter lvl 7
    How huge is the difference in terms of quality..if there is one.
  • AlecMoody
    Offline / Send Message
    AlecMoody ngon master
    My understanding is that there isn't significant difference in quality between agisoft and RC if you are using comparable settings- just a massive difference in time and the workflow in reality capture is a lot easier. That said, if this scan takes 3 hours to process in RC but 30 hours to process in agisoft, agisoft is a non starter and I would be forced to process on much lower settings to get it done. I haven't done any A/B testing as I have no interest in making my computer unusable for days at a time.
  • RyanB
    I'm currently using VisualFSM and Meshlab to learn photogrammetry techniques.  Your results are much better than mine so far. 
  • AlecMoody
    Offline / Send Message
    AlecMoody ngon master
    This is a similar workflow on a hard surface part. In this case the low is an optimized version of zremesher and not decimation.

  • Mant1k0re
    Offline / Send Message
    Mant1k0re polycounter lvl 8
    O.O Wow. Any chance for wires for the low on this last example?
  • DaveW
    Offline / Send Message
    DaveW polycounter lvl 14
    One workflow I haven't been able to figure out, at least from my own experimentation, is how to do a scan involving two sides of an object. The photo-scanning process is relatively simple for rocks, where you don't care about the underside, but for that hard surface model presumably some part of it was laid flat against your steel plate.

    My attempt at a solution was to use an evenly-lit white background for the object, so that the software was only able to compare points on the object (and therefore I could flip the object over, and it would create the underside without any issues). Wasn't very successful since you're removing a lot of the tracking data.
  • AlecMoody
    Offline / Send Message
    AlecMoody ngon master
    For that upright I suspended the model with rope/zip ties. I also used rare earth magnets to rigidly attach the model at a few points so it wouldn't sway. I am planning to build a more elaborate fixturing setup.



  • armagon
    Offline / Send Message
    armagon polycounter lvl 11
    I'm curious about your lighting setup. Do you have more pictures of it?
Sign In or Register to comment.