By the way is this all NDA? How much can I post on facebook? 
cool capture.
so what do you think of fathom so far? 
okay - so here is the difference between me, and say - autodesk: i dont have anyone to be accountable to - but i do have to concern myself with competition.
however, i like transparency. OTOH, i dont want to over promise and under deliver…
SO - i’ll make a deal with you guys - as long as you swing it by me for approval first, and i agree - and you credit thinkbox software/fathom then i will be happy.
What do you want to post? that capture above?
cb
Well my first impressions is that for capturing a “volume” live it’s not actually that well suited. The range is too short and the FOV is too narrow. So even to capture a relatively small space like our main studio that was something like 20 Kinect shots and it’s still pretty spotty. In order to cover a stage it seems like you would need hundreds of Kinects and I’m not entirely certain how you would keep them within about 20 feet at all times of every subject. It would be at the limit of their range if suspended from the ceiling.
I just started experimenting with capturing one person but even one person at a decent resolution would require 4 kinects… minimum. And once you get to 4 kinects it’s going to be a big challenge on your part to perfectly characterize the lens distortion in order to seamlessly merge the 4 kinect’s data.
Once you get SLAM going like the KinectFusion demo I think it would be great for capturing sets since you could wander around and then import the 3D model into PF-Track or Boujou etc and really improve your solves. It would also be great for lighting integration. Mount a 5D next to a kinect on like a fig rig and wander around capturing a 3-exposure (18-20 stop) image every second and then project that onto the SLAM derived model, turn on brute force GI and you could light your 3D exactly as it was on set, or at least use it for reference to place your CG lights and use the Kinect/5D scene for your reflections.
The challenge though is that none of this really seems to work very well outside so it would be limited to interiors.
I would be interested to see if this could solve camera motion for those really hard shots where it’s difficult to place tracking markers. A nice new feature of PF-Track is that you can provide a “hint” animation track. I think it could be more than accurate enough to get the general motion for the hint-track. Again though, range seems like it would be the largest issue if you need to track a scene larger than about 20 feet. For greenscreen that would be problematic since of course you want to push your talent as far away as possible to avoid spill.
I do really think if you could make a camera/kinect/laptop combo for real-estate agents it would be great as a Unity Engine online walkthrough tool.
Oh, one thing that it would be really really awesome for is some sort of super stylized music video look. Particles generating particles and all that. So some really stylized animation could be cool.
Yeah we were thinking about throwing that up on our Straightface Facebook as a little teaser of “cool stuff we’re experimenting with”. Although we would probably re-do it with a tripod and more care in aligning the images if we did.
thanks for the feedback.
i dont necessarily agree that it can’t be used to capture a volume - my feeling is this
- you lidar the entire location in advance
- you use an array of kinect’s to caputre the volume at 24fps or 30fps or whatever.
- you match points and locations ot the lidar and thus you have a reasonably accurate measurement of what occurred in the shoot. the 1mm or whatever lidar scan gives you absolute positioning, and the kinect gives you relative [and less accurate] positioning, but surely enough for many cases.
once v2 of the kinect technology hits the market, many of these problems hopefully get reduced
cb
I wonder if you could hack the kinect to up the voltage on the IR projector bulb…
please, try. i hear people have hacked all sorts of things! in fact, if you are electrically/mechanically inclined let’s talk further.