Adding roundness to 2D to 3D conversion.

[Stefan Christou]

Hi guys,

I'm doing a project on 2D to 3D conversion using Smoke for Mac as part of my Masters in Broadcast Futures at Ravensbourne College her in London, Yooooookayyy.

Anyway: I'm having a bit of trouble figuring out how on earth to add roundness to a roto'd layer using wireframe models.

I'm using some of the R3Ds from the MX reel. (I hope this is OK).

Does anyone have any ideas on this? Even if you have some tips from a different application that would be great. s3D is pretty new in smoke, and there doesn't seem to be too much info on 2D to 3D in smoke.

Anyway I'd really appreciate the help on this one. Thanks in advance and I hope y'll are having a good weekend.

If anyone is interested you can keep up with my progress on http://scraveblog.wordpress.com.

Stefan

 

[Stefan Christou]

Here's a look at how I'm getting on so far...

http://www.youtube.com/watch?v=GOe58Xct58Y&hd=1

 

[Fabricio Morato]

Hi,

how are you rotoscoping the car? by parts?

I've never used smoke but used nuke, and theres a few ways you could do it inside nuke. You can go the 3d way or 2d.
I would rotoscope each part of the car that has dimension individually, and make it a depth map (grayscale image) white being whats in front and black whats behind (it doesn't really matter because you can invert the values).

After having this you can apply a 3d displacement to the image or use 2d displacement or do it hybrid.
Note that you may need to create a clean black plate for one of the eyes...

and for round corners you can apply individually to each roto an edge feather or something like it, and with the displacement it will make it rounded...

I think that in the imagineer systems website theres a tutorial on 2d to 3d conversion using mocha and nuke you can check it out...

hope this helps

 

[M Most]

After having this you can apply a 3d displacement to the image or use 2d displacement or do it hybrid.

Displacement maps are one way. A better way is to create a 3D model of the object (match moved if necessary) and projection map the image back on to it. That gives you true 3D depth that the stereo cameras in Action can understand. You can also use the 3D model to directly derive a depth map if you want to work that way. Of course, it also requires a lot more work.

 

[Stefan Christou]

Thanks for the advice guys. Projection mapping sounds like the way to go. Since I'm doing this as a uni project, I'll have a pretty detailed write up of it that I will post so anyone can follow my steps.

Stefan

 

[Stefan Christou]

So... I finished this a few weeks back, I didn't get to add a Z-map, as Smoke for Mac does not yet support this feature. However, I think I have got some pretty good results, especially to the untrained eye.

Let me know what you think guys, and also check out the workflow document if you want to have a look at repeating my experiment.

Stefan

http://vimeo.com/20533090

http://stefanchristou.com/​2dto3d/​3d-conversion-how-to.pdf