Agisoft Metashape

Agisoft Metashape => General => Topic started by: chadfx on July 22, 2013, 07:39:50 PM

Title: Scene Reconstruction from High Spatio-Angular Resolution Light Fields
Post by: chadfx on July 22, 2013, 07:39:50 PM

A research project from Disney Zurich for those interested...

http://www.disneyresearch.com/project/lightfields/
Title: Re: Scene Reconstruction from High Spatio-Angular Resolution Light Fields
Post by: meshmaster on July 22, 2013, 10:36:05 PM
Sweet... awesome research!

Cheers,

J
Title: Re: Scene Reconstruction from High Spatio-Angular Resolution Light Fields
Post by: chadfx on July 22, 2013, 10:41:37 PM
Yeah, they've got some smart people over there...really clever solutions to lots of other problems (VFX'ers can relate to a number of these!) and they seem to continue to try improvements and refinements to other existing solutions.

Some slightly off topic viewing for you:

http://www.youtube.com/user/DisneyResearchHub/videos
Title: Re: Scene Reconstruction from High Spatio-Angular Resolution Light Fields
Post by: Wishgranter on July 22, 2013, 11:25:13 PM
The only drawback is that their solution is and highly probably will be not for "general" public, they release it not outside of lab..... 
Title: Re: Scene Reconstruction from High Spatio-Angular Resolution Light Fields
Post by: jedfrechette on July 23, 2013, 02:23:07 AM
Quote from: Wishgranter on July 22, 2013, 11:25:13 PM
The only drawback is that their solution is and highly probably will be not for "general" public, they release it not outside of lab.....

I wouldn't necessarily assume that. Disney has a pretty strong history or releasing internal tools as open source.

http://www.disneyanimation.com/technology/opensource

The big VFX & animation houses like Disney also play a pretty important role in driving feature development in commercial software like Mari & Nuke, even when those packages started out as in-house tools.

If this method proves useful in production I would not be surprised at all to see it pushed out to either third party commercial software or potentially even an open-source library.

Regardless, although I only skimmed through the paper it would seem that it contains all the detail needed for a third party developer to write their own version of the algorithm.