This site relies heavily on Javascript. You should enable it if you want the full experience. Learn more.
Projector node tutorial using ReconstructMe
Complete, just need to make a video and package up everything. Any volunteers to test? Need to track down missing modules, etc that may be scattered in my repo
@mrboni -
multi projector is possible with tweaks / a little hacking.
it's not meant to be a system, more the components and an example of how to use it
no reason why not multi-projector though
Thanks to any advanced users who can check through this!
I'm flying to korea tomorrow so will pick up on the comments in a couple of days, i hope it's not totally broken on first try :(
checking
first coup of things i think reconstuctme outputs .stl, and size difference because X is binary and reconstuctme outputs TEXT, also UV-ing it is mental hehe
hit 'ok and hide'
it's a divide by zero bug somewhere in the core (this is mentioned in the video also, it's to do with an empty spread somewhere that disappears after you've calibrated)
the alpha will give you error messages that are hidden in the release (beta) version of vvvv
you might also need to get the latest alpha (28) if you encounter other bugs
@antokhio : reconstructme can output 3ds, stl and (i think) ply.
as mentioned in the video, choosing the extension of the filename determines which format of file it outputs
i know the video is a bit long, but please check through it if you've got any questions about how each stage works, you can see there how to convert the 3ds into x for vvvv
the tool to do so is in the zip (just drag the 3ds file onto it and wait)
@elliotwoods, thx, i will check it.
@antokhio this is usual error, that appears on some of our openCV plugins since latest alpha build. I think on next week it should be fixed in my fork, or Elliot can fix it early.
waiting for the darkness to come, as i remember pade's example want to check this one with projector
what bugs me is a perspective, since vvvv have linear perspective and lens have also barrel distortion witch is unnecessary with long throw optics but might be bit bitchy with short-throw
Personally i would give you a Nobel prize for that because it took 2 times less time then usual to do that.
So, reporting back from my findings:
1-t bit tweaked mesh select so it outputs points to projector so you can see bit what happening.
2-d tommo i gonna give it a test with less complicated geometry so we can see more accurate result
3-d tab don't seems to accept Ctr hotkey (to change value 0.0010 instead of 0.0100), also would be nice to have some step offset on key say + -
rest is super cool, camera preview shows really close match to the position of my projector.
also a note a used ms-sdk kinect
Wow! Thats great
Really appreciate you posting the images
CalibrateProjector (i.e. CalibrateCamera) can calculate lens distortion also.
To render it you'd need to either a custom Vertex shader and dense meshes on all your objects, or an extra renderpass pixel shader and lose your sharp edges (probably still fine though).
So if people are finding barrel distortion issues, then we can look at making that pixelshader for lens distortion correction. You'll just need to add a couple more points into the dataset (e.g. 7 minimum).
Cheers for the notes on Table nodes. Will look into that.
Any other comments on Table nodes are also really welcome.
Other note: this routine is pretty similar to Kyle McDonald's
'mapamok' so there are cross platform equivalents out there.
my name is Christoph and I'm the main developer of ReconstructMe. I've seen this blog popping up recently on my monitoring and I'm really interested in the software and how ReconstructMe can help to simplify processes.
Here's my idea: Elliot, I'd love to put this onto our gallery (reconstructme.net). In case you are interested, could you write 2-3 sentences describing your system and how ReconstructMe helps speeding up things? Some screenshots and video footage describing the process would be ideal. Just drop me an E-Mail at
hmm seems interesting the old projector node was good. what's new with this one?
I am of course up for testing it out.
+1
Is it on GitHub?
@alg all the nodes are on github
my 'subrepos' branch + Image repo
and the tutorial files are at:
https://github.com/elliotwoods/VVVV.Mapping.Tutorials.Mapping2-3D
am packaging now for testing
Multi projector support?
@mrboni -
multi projector is possible with tweaks / a little hacking.
it's not meant to be a system, more the components and an example of how to use it
no reason why not multi-projector though
Ok right.
So here's a release for the testers.
You'll want to make sure VVVV isn't using any previous version of the OpenCV nodes that you have.
This is for the testers only, please don't expect any support or to use these files in your next project ;)
Notes:
Thanks to any advanced users who can check through this!
I'm flying to korea tomorrow so will pick up on the comments in a couple of days, i hope it's not totally broken on first try :(
download:http://www.kimchiandchips.com/files/Projector%20node%20test%20release.zip
legend
checking
first coup of things i think reconstuctme outputs .stl, and size difference because X is binary and reconstuctme outputs TEXT, also UV-ing it is mental hehe
missing intrinsics/extrinsics nodes/links in the calibrateprojector(openCV) module...So cant solve anything. Or it is just me?
Ok, forgot to remove all previous openCV stuffs. solved!
on alpha 27.2 having constant error looks like that:
hit 'ok and hide'
it's a divide by zero bug somewhere in the core (this is mentioned in the video also, it's to do with an empty spread somewhere that disappears after you've calibrated)
the alpha will give you error messages that are hidden in the release (beta) version of vvvv
you might also need to get the latest alpha (28) if you encounter other bugs
@antokhio : reconstructme can output 3ds, stl and (i think) ply.
as mentioned in the video, choosing the extension of the filename determines which format of file it outputs
i know the video is a bit long, but please check through it if you've got any questions about how each stage works, you can see there how to convert the 3ds into x for vvvv
the tool to do so is in the zip (just drag the 3ds file onto it and wait)
working like a charm :)
beta 27.1
brilliant!
@elliotwoods, thx, i will check it.
@antokhio this is usual error, that appears on some of our openCV plugins since latest alpha build. I think on next week it should be fixed in my fork, or Elliot can fix it early.
waiting for the darkness to come, as i remember pade's example want to check this one with projector
what bugs me is a perspective, since vvvv have linear perspective and lens have also barrel distortion witch is unnecessary with long throw optics but might be bit bitchy with short-throw
maaan this is gonna make mad-mapper apple hipsters soooo jealous:D
Personally i would give you a Nobel prize for that because it took 2 times less time then usual to do that.
So, reporting back from my findings:
1-t bit tweaked mesh select so it outputs points to projector so you can see bit what happening.
2-d tommo i gonna give it a test with less complicated geometry so we can see more accurate result
3-d tab don't seems to accept Ctr hotkey (to change value 0.0010 instead of 0.0100), also would be nice to have some step offset on key say + -
rest is super cool, camera preview shows really close match to the position of my projector.
also a note a used ms-sdk kinect
Wow! Thats great
Really appreciate you posting the images
CalibrateProjector (i.e. CalibrateCamera) can calculate lens distortion also.
To render it you'd need to either a custom Vertex shader and dense meshes on all your objects, or an extra renderpass pixel shader and lose your sharp edges (probably still fine though).
So if people are finding barrel distortion issues, then we can look at making that pixelshader for lens distortion correction. You'll just need to add a couple more points into the dataset (e.g. 7 minimum).
Cheers for the notes on Table nodes. Will look into that.
Any other comments on Table nodes are also really welcome.
Other note: this routine is pretty similar to Kyle McDonald's
'mapamok' so there are cross platform equivalents out there.
@ elliot
Do you have any edge blending techniques working with this setup?
My test:
http://bit.ly/KsODW6
About 30mins for assemble scene, hardware, scan and calibrate
http://youtu.be/vTMejd6GJS8 ghm
Hi,
my name is Christoph and I'm the main developer of ReconstructMe. I've seen this blog popping up recently on my monitoring and I'm really interested in the software and how ReconstructMe can help to simplify processes.
Here's my idea: Elliot, I'd love to put this onto our gallery (reconstructme.net). In case you are interested, could you write 2-3 sentences describing your system and how ReconstructMe helps speeding up things? Some screenshots and video footage describing the process would be ideal. Just drop me an E-Mail at
info@reconstructme.net
Best regards and thanks in advance,
Christoph