| View previous topic :: View next topic |
| Author |
Message |
grabiller
Joined: 08 May 2012 Posts: 4
|
Posted: Tue May 08, 2012 8:40 am Post subject: Deep Datas Extraction (dtex,EXR2,..) |
|
|
Hi,
Using SynthEyes and Nuke, I wonder if something is planned for us to be able to extract deep datas from solved shot(s).
Now that deep compositing is available in Nuke, I believe it would be wonderfull to have this feature in SynthEyes.
I understand deep datas usualy represent several samples per pixel from CGI rendered images, and that we can allready extract depth passes from SynthEyes but I was thinking that with possibly several solved shots, SynthEyes may be able to provide several samples per pixels as well. But even 1 sample/pixel would be a start.
Any word on this would be appreciated.
Thanks in advance.
Cheers,
Guy.
--
guy rabiller | radfac founder / ceo |
|
| Back to top |
|
 |
GraphicsKid
Joined: 25 May 2009 Posts: 480
|
Posted: Tue May 08, 2012 2:47 pm Post subject: |
|
|
| Samples/pixel of... what? The set reconstruction? |
|
| Back to top |
|
 |
borgus
Joined: 30 Dec 2009 Posts: 144 Location: bremen, germany
|
Posted: Wed May 09, 2012 2:54 am Post subject: |
|
|
Hi grabiller,
deep compositing seems to be a very useful thing, but I don't see what
kind of data SE could provide to a compositing package. Are you talkig
about motion vectors of pixels?
I'd love to know more...
Seb |
|
| Back to top |
|
 |
grabiller
Joined: 08 May 2012 Posts: 4
|
Posted: Wed May 09, 2012 5:24 pm Post subject: |
|
|
Hi,
Well, once a shot is solved, theoricaly you could add a tracker at each pixel and then get the 3D position at each pixel, right ?
I'm no expert at the math involved behind tracking but I have the feeling that it could be possible - theoricaly - to directly extract the actual depth information at each pixel without the need to manualy reconstruct the scene with meshes, once a shot is solved.
This may require long processing time of course but it's no big deal compared to the value of the datas we could get from this.
Then depending on the math precision of course, I don't see why it couldn't be possible to retrieve this information on a sub-pixel level.
A proof of concept could be done by manualy (or better through scripting of course) adding a tracker per pixel. I'm not sure if SynthEyes could actualy manage as much as trackers, but I believe a specific internal process created for this purpose could do the job.
I would be interrested to hear from Russ on this matter.
Cheers,
Guy.
--
guy rabiller | radfac founder / ceo |
|
| Back to top |
|
 |
grabiller
Joined: 08 May 2012 Posts: 4
|
Posted: Wed May 09, 2012 5:29 pm Post subject: |
|
|
| borgus wrote: | Hi grabiller,
deep compositing seems to be a very useful thing, but I don't see what
kind of data SE could provide to a compositing package. Are you talkig
about motion vectors of pixels?
I'd love to know more...
Seb |
Currently and AFAIK, only Nuke support deep compositing and only through the dtex format from Pixar/prman. Deep datas are actualy very similar to deep shadows datas.
Upcoming OpenEXR 2 should also support deep datas. Houdini has it's own format but not natively supported by Nuke albeit It seems a few peoples are working on some plugins.
Cheers,
Guy.
--
guy rabiller | radfac founder / ceo |
|
| Back to top |
|
 |
|