AutoDesk PhotoFly - 123D Catch

[Stan Zaslavsky]

Hi there,

 

Came across this tool that is now called 123D Catch.

 

Has anyone used it effectively and is it useful for larger models (ie statues)?

[Jonathan]

Hi Stan,

I've experimented with Photofly/123D quite extensively now. On the plus side, it is free, very fast (as a cloud processor) and easy to use. It is limited however in the number of photos that can be uploaded AND in the final model's polygon count and therefore detail, quality and usefulness. A maximum mesh count of 322,000 tri's seems to be about the max no matter the complexity or amount of photographs. I recently modelled a small statue and used both 123D and Agisoft Photoscan (a commercial application). http://pixogram.co.uk/?page_id=36 With a raw model of 8 million tri's compared with 123D's 322,000 you can imagine the difference. I still however use 123D for texture detail projected onto higher mesh models and for quick survey/reference models of buildings and even streets.

[Stan Zaslavsky]

Hi Jonathan,

Thank you for your detailed input - I'll check out Agisoft - as we may need the higher resolution for this particular model.

 

Is there any good tutorials that you have found for 123D or was it all just trial and error?

[Jonathan]

Hi Stan,

Its very much trial and error really. A decent understanding of photogrammetry and photography techniques would certainly help. The reality capture is one thing, dealing with a highly complex and dirty mesh is another and is extremely project/view/animation dependant. Max 2012's retopology tools and '0' Render to texture are a godsend and a sculpting package such as Mudbox is essential. Also look at MeshLab for algorithmic cleaning. For me it has been a gradual learning curve but a fun one!

[danulrich]

Hi Stan,

 

i've done some testing with it, and its quite ok, but smaller details need closeup photgrpahs, just as Jonathan said, its limited to something like 300k tris... on the other and, with laser and sl scanners you have to scan smaller areas too, and align them afterwards in rapidform, meshlab or whatever has a fairly decent ICP algorythm...

 

here are some scultpures i'm working on currently, but i havent been able to work further on them for 2 weeks now cause the crappy weather here...

 

http://www.flickr.com/photos/danulrich/6765257793/in/set-72157629001361761/lightbox/

[Tommy L]

Has anyone found a commercial use for these kinds of photogrammetry or scanning applications? I use scanning for some projects, but its just for capturing objects where the capture/texture process is quicker using a scanner than traditional modeling/mapping. Yet to find a dedicated market.

[Jonathan]

Tom,

I've found some commercial application for reality capture in the heritage arena. Either as a digital reconstruction tool (for virtual restoration) or as an accurate 3D model prototyping tool prior to the construction of real prototypes before manufacture. The possibility for 3D printing could also integrate with the above mentioned as well as having a use in public outreach, education etc etc. A great new example is CyArk's laser scanned Mount Rushmore project: http://archive.cyark.org/mount-rushmore-digitally-reproduced-blog I think its really a case of making markets at the moment...

[ziemowitb]

Hi all,

 

Do you actually know what kind of matching algorithm they are using ??

 

regards

[Bruce Hart]

We've used it to model people's heads in our office, but nothing actually work related It does a surprisingly good job, but you need to give it a lot of photos (say 20+) for it to work well.

[Simon Moir]

I think it's pretty cool. There's a free (for now) program called Mesh Mixer that's great for cleaning up the resulting mesh.

 

[ATTACH=CONFIG]47450[/ATTACH]

[Jonathan]

Mesh Mixer looks pretty cool! I'd like to see it tackle a 15million poly model though. Another alternative is MeshLab which is algorithmically based. Sculptris another free tool is good for cleaning up small, light models but no comparison to Mudbox or ZBrush.