Immersive 3D Video

Playback of a 3D video stream of a dance performance in an immersive CAVE environment at the UC Davis KeckCAVES facility ( http://www.keckcaves.org ).

The 3D video data was provided by the CITRIS Tele-Immersion lab at UC Berkeley ( http://tele-immersion.citris-uc.org )

The first half of the video was shot from a fixed camera position outside of the CAVE; the second half was shot hand-held.

Category:

Film & Animation

Tags:

• CAVE
• VR
• immersive
• 3D
• video
• dance
• performance
• teleimmersion
• CITRIS

[nightmakerrecords]

@sadunkal your a fucking dickehead.

[spelunkerucd]

@nightmakerrecords No, on the contrary, he's very insightful. Check the comment he made a year ago about Project Natal, and compare to our Kinect videos to see how right he was.

[pigsareit]

@p4tr1ck77 I imagine a recording of a someone passed on, and then sitting in the same room with them, as they talk to you. Maybe not a conversation, but the idea is still pretty awesome.

[p4tr1ck77]

This is amazing, just think of the possibilities.

Talking to your son whos halfway around the world, and hes right infront of you.

Think of the video gaming experiences.

Awesome for education....show your students exactly what you are talking about, like its right in front of them

I love technology...

[Outsider0210]

is it fake ?

[sadunkal]

Maybe all you need is an Xbox360. Check out "Project Natal"... It surprised me. It's not yet exactly clear how precise it is though.

[spelunkerucd]

That's the one. Christian presented the method to my group a few months back, and I wanted to start using it immediately, but unfortunately he told me then that it took quite a bit of processing between capture and playback. So I shelved that idea for the time being. :(

[sadunkal]

Ow alright. I think I just found the "very impressive method" you're talking about; "Performance Capture from Sparse Multi-view Video". It's very impressive indeed, especially since it's done just with 8 cameras. And I can understand how it's not suitable for real-time, yet. I was imagining a lower quality motion capture and a less detailed model/texture when I made the above comments. :)

[sadunkal]

Sorry the paper is only cited by Theobalt, not authored by. It's actually from Aachen University, by Hornung and Sar-Dessai.

[sadunkal]

Oh, ok. :) Keep working on it, it's the future...

And when you say that "it's not real-time yet" are you referring just to the modeling or does also the motion capturing require time to process?

...

I just noticed a paper by Theobalt et al. from 2005 where they "present a self-calibrating framework for optical motion capture, enabling the reconstruction and tracking of arbitrary articulated objects in real-time". You mean there are still problems with real-time motion capture of humans?