Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Tonaci's HypeVR

Super badass! I really want to try it out. But... and I hate to always be captain buzzkill but the math is WAY off on "84k". By that figuring the RED Dragon is a 6,000k camera if you shot one row of pixels and counted all of the other rows as columns. If you had 0 pixels of overlap you would at best have:

The correct math for horizontal resolution of the stitched image is: Sqrt(14) = 3.74 cameras of horizontal extra coverage * 6k = 22k.
However on average for panoramic stitching you're most likely each camera will only contribute ((XCams - 2)*0.33)+2) = ((3.74-2)*.33)+2 = 2.54 * 6k = 15k of unique horizontal resolution. Which is plenty for current applications since most VR headsets have at best 2.5k horizontal resolution and 100 degrees of FOV. 3.6 * 2.5k = 9k ideal.

I'm curious how they're retiming the lidar data since it's at 10hz and the cameras are presumably at 60+hz. Also the lidar is only about 400x200 points at 10hz. That's around 1 horizontal voxel per degree. So I assume there is some sort of synthesis of using the stereo disparity maps for moving objects like people and the lidar is just for static scenery?
 
I'm usually the guy who points out the mathematical fail. ;)

Besides, we all know that they just did 14 * 6144 / 1024 = 84.24... 84K!!! WooHoo!. Heheheh.



That said, Tonaci's rig looks totally badass. I want to play with it. I bet it's fun, judging by the grin on Justin's face.
 
Very cool. Tonaci is working hard all year long.

@Jeff - There's overlap and I hope it's in the 20-33% range :)
 
Phil Holland said:
Very cool. Tonaci is working hard all year long.

@Jeff - There's overlap and I hope it's in the 20-33% range :)
Click to expand...

Yes, I know there's overlap. We were just commenting on the 84K claim in the other write-up. :) I was just saying I know where they got the 84K number from. I'm pretty much in agreement with Gavin on this with the approximate 15K of final unique horizontal resolution.
 
Pretty cutting edge.... Tonic, I hope you keep us up to date on it all.... I believe VR is a big part of the future (compared to 3D) and I would love to learn more...
 
Sigma's! Interesting. Using EF AutoFocus to pull them all? Used to be able to do this with GigE cable on two MX brains for stereo on older v3 something firmware with Denecke-feeding Genlock but last time I tried I couldn't get it to work (though I didn't try that hard to find a solution).

Automated stitching is a big one. Well done Tonaci!

Would absolutely love to see some of this VR footage on a Rift. I had a go at the Go Pano 360 Lens. Darn that footage looked terrible : )

Craig Lees
ED UK
3dstories.co.uk
 
Lakis Amarantithis said:
Is the final product 1 or 7 files in sync?
Click to expand...
Should not be too difficult to stich together a double half sphere canvas to get something about 220 x 360 degrees stereoscopic motion quicktime VR. then compress it and use those as realtime rendered pan and tilt Oculus Rift source material.
 
how many media managers does that thing take with check sums? can i get a dozen trash cans over here, stat lol?
 
I helped Luis Flores with some of my cameras for a rig very similar to this about a year ago, i think he was shooting the dallas cowboys with it, is this the say thing? he showed me some footage about 5 months ago it was sick!
 
Gavin Greenwalt said:
Super badass! I really want to try it out. But... and I hate to always be captain buzzkill but the math is WAY off on "84k". By that figuring the RED Dragon is a 6,000k camera if you shot one row of pixels and counted all of the other rows as columns. If you had 0 pixels of overlap you would at best have:

The correct math for horizontal resolution of the stitched image is: Sqrt(14) = 3.74 cameras of horizontal extra coverage * 6k = 22k.
However on average for panoramic stitching you're most likely each camera will only contribute ((XCams - 2)*0.33)+2) = ((3.74-2)*.33)+2 = 2.54 * 6k = 15k of unique horizontal resolution. Which is plenty for current applications since most VR headsets have at best 2.5k horizontal resolution and 100 degrees of FOV. 3.6 * 2.5k = 9k ideal.

I'm curious how they're retiming the lidar data since it's at 10hz and the cameras are presumably at 60+hz. Also the lidar is only about 400x200 points at 10hz. That's around 1 horizontal voxel per degree. So I assume there is some sort of synthesis of using the stereo disparity maps for moving objects like people and the lidar is just for static scenery?
Click to expand...

Did Wesley Crusher just show up? Too funny!

Just kidding, but seriously, cool stuff, Tonaci!
 
Looks awesome.

Why are you guys using different lenses on some cameras? They are not all the same by the looks of it which would lead to problems with the stitch possibly?
 
Great
 
You must log in or register to reply here.
Back
Top