Re: Can you use multiple Kinects at once?

[Nicolas Burrus]

Hi,

Currently only one sensor is supported. While we did some custom
projects with several sensors, multiple sensors is not on our short
term roadmap for Skanect. But this may change if there is enough
interest!

Cheers,
Nicolas

On Mon, Feb 4, 2013 at 7:04 AM, Matt Moo <asdf...@gmail.com> wrote:
> Lets say I want to build a booth to scan an object. If I had 4 Kinects
> plugged, will the software recognize all 4 angles in 1 shot, or will it have
> me pick which Kinect to read from (ie: only using 1 kinect)
>
> --
> You received this message because you are subscribed to the Google Groups
> "Skanect" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to skanect+u...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[James Allan]

This is something I would like to see implemented. It would help speed up the scanning process and maybe even allow for 'instant' capture - I'm thinking those hard to hold poses, etc. It would perhaps remove some of the errors that inevitably creep in when some is trying to hold still for 20 seconds or more.

[Nicolas Burrus]

We hear you. This is indeed becoming a common request, let's see if we
can find an easy way to integrate this.

Thanks for the feedback,
Nicolas

[James Allan]

Put me down for beta testing! :D

[metrogee...@gmail.com]

I will be interested too as I have 3 kinects with me.

[Ian Shook]

I'd like to see this too!

[David Moore]

I haven't seen anything recently, and I am new to 3D scanning. Has multiple scanners been implemented, or is it still not on the short term roadmap?

Thanks,

David

[Nicolas Burrus]

Still not in the roadmap, sorry.

[tapej...@gmail.com]

All I can see is several people getting in eachother's way and ending up scanning eachother as well as the subject. More than one camera would definitely be useful for MoCap but for model reconstruction I think it would create more problems than it would solve.

 

[r...@crowtreesstudio.com]

Multiple sensors would be a god send! (you hear that Nicolas? ... I just called you a GOD!! <];) ). Set them up around your subject and snap a shot instead of walking around them or spinning them on a turn table (or office chair). Please scribble this onto the road map. Or record a sequence of movement and pull out the still frame that you need. Oh man ... goosebumps!!

PS

Radeon support would also be nice :) just sayin.

[Nicolas Tisserand]

Roger that!

We heard the multiple sensor requests (you're not the only one asking
about it) loud and clear!

Also noted about Radeon GPU acceleration.

Thanks!

[James Allan]

re: Radeon GPU acceleration

Pretty please :D

[Donato Cannatello]

May be multiple sensors could give a  boost to quality of scan and holes in the resulting mesh ?

I think this can be the only step to bring this system to a professional level.

[Bernhard Spanlang]

Hello,

I am also very interested in multiple sensor support.

Is there any news on this ?

Thanks,

Bernhard  

[Sungwook Su]

I figured out it is almost impossible to take whole body scan with one sensor. 

especially lag has a lot of holes.

plz guid me better way to take whole body scan.

thanks

Su

[CÉ3D 97]

Up!! 

I 'm also very interested by scanning with multiple sensor!!

[We3DU]

Thought I'd add my voice to the throng. :-)

There's at least one player in the space who already supports 2 Kinect sensors and actually up to four sensors under certain circumstances. It would certainly be a very helpful feature when scanning people standing up. In fact, I'd probably be prepared to buy some kind of multiplexer if that was the only way. An application for a Raspberry Pi, maybe??

My ten cents.

A

[Vitaly Vyazovsky]

Right now you can use multiple instances of scanect, acquire  meshes simultaneously, and then combine them using external software - say, meshlab.
Tried this on my macbook pro with kinect and primesense, worked perfectly. I had to use version 1.5 though (later versions doesn't allow multiple instances)
http://www.cultofmac.com/200062/run-multiple-copies-of-an-app-at-once-on-your-mac-os-x-tips/
With one instance set to CPU and second to GPU I was able to get a descent FPS. 

четверг, 25 сентября 2014 г., 9:06:10 UTC+4 пользователь Sungwook Su написал:

[Christian]

Support for multiple Kinect/Xtion sensors.
+1

Idea would be to have 3-4 sensors mounted with different high.
Beeing able to capture full body scan (with no holes) in one 360 turn.

[Kiwi Tsai]

So does the multiple sensor support on version 1.7 now?

I also need to scan faster and good quality.

Christian於 2014年10月3日星期五UTC+8下午7時57分00秒寫道：

[Nicolas Burrus]

No, multiple sensors are not supported in v1.7. We could easily enable
multiple instances of Skanect to enable Vitaly's setup though. Maybe
with a command line options, since this is more for advanced users.

> --
> You received this message because you are subscribed to the Google Groups
> "Skanect" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to skanect+u...@googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.

[eRoom-D]

I would love to see that happen, even as an alpha/development option. Using something like Sandboxie is more problematic that helpful.

David

You received this message because you are subscribed to a topic in the Google Groups "Skanect" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/skanect/hSJg70scXwk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to skanect+u...@googlegroups.com.

[Matt Kimball]

Bump bump

+1 to multiple sensors.  Maybe they could align to each other in a prescan setup, done off an inverted cube, carefully marked.   Similar to how David Laserscanner orients the camera to the subject area?  It wouldn't surprise me if OpenCV made this fairly simple but I don't know for sure.

[Borko Jovanovic]

Another vote for multiple sensors.

Full-body scans with 3 sensors @ different heights. one turn.

Something like those guys explained: http://www.math.zju.edu.cn/ligangliu/CAGD/Projects/Kinects-CapturingHumans/

[Christian]

+1 for multiple sensor support

[Tobby Ryan]

+1 also

I am happy that both kinects capture during my stupidity I left 2 connected.  both folders were populated and I could "disable" one by moving and doing an offline reconstruction, as normally it would merge them both at the same time, then the same with the other.  Post merging in Meshlab = not too bad.

it would be nice to just have a standard stationary rig with 2 or 3 kinects instead of moving one up/down while person is rotating.

so I definitely vote +1

[falconer odland]

+1 for this multiple sensors. Like others have mentioned this will be a huge step towards instant capture of hard to hold poses.

[StarManta]

Adding in for this as well. Multiple sensor support would be worth a lot to me. I already have a Skanect license, and I for one would be happy to fork over more for multiple sensor support.

[Michelle Folkman]

Ditto. What everyone has said

[Tobby Ryan]

*prays that this is somewhere up on the to do list lol.

I would soooo like multiple sensor support. 

[Sebastian Galli Versteegh]

+1 to multiple scanners support. I'm a Senior software engineer/Solutions Architect, please let me share with you my humble opinion:

I think that the trick to build this feature could be (2 different approaches):

Proposal 1). The first consist into put all devices in the same angle of vision (all devices are generating similar information) [Synchronization phase], after this the user should move one camera to the opposite side (180º if there are only 2 devices, 120º if there are 3 devices and so on) .

Proposal 2). The second one is detect those key points from the multiple views to be able to synchronize all views. How we could achieve it? -Rotate the object some degrees at least until receive the necessary information from source A to be able to compare it with the information received from source B to looking for those key points (same real points from different views), with this information should be enough to calibrate position/distance of each sources. 

Of course you knew it, but just in case.

Best regards,

Sebastian.

https://es.linkedin.com/pub/sebastian-galli-versteegh/a/ab0/796

[Tobby Ryan]

Any updates on thoughts or plans to support 2 sensors without having to take each stream and reconstruct offline separately and mesh mix ?
Or is the focus going to be more on the structureio device sdk?

Any other tips/tricks other than just the 80% start near the top for the bounding box config.ini ?  Perhaps an advanced listing of maybe command line things to tinker with?

[Nicolas Burrus]

Hi,

More flexibility for the bounding box size and position is definitely something we want to add in the near future.

However even if we understand that it would be very useful for some specific use cases with custom rigs, full multiple sensor support is not likely to happen soon, we are indeed focusing more on making the single-sensor scanning experience itself more robust and higher quality.

Note that recent versions of Skanect allow multiple running instances though, so even if not ideal, the workflow described by Vitaly should work.

Thanks again for all the feedback,

Nicolas

--

[Tobby Ryan]

So basically.
1.    Scan with 2 kinect

2.    go to the skn folder and under images, create a folder and move the -1 into it.
This allows me to only process the -0 folder
load the skn again to do an offline reconstruction of the -0 data only.
fill,color, save as model0.ply

3.   go to the skn folder and under images copy.paste the -1 back out, and move the -0 folder inside it
This allows me to only process the -1 folder
load the skn again to do an offline reconstruction of the -1 data only
fil, color, save as model1.ply

Load both model0.ply and model1.ply into meshmixer and mix the meshes
export as finished.ply

yes i skipped the actual steps on the mixing process..

[David Moore]

Nicolas,

You mention that you can run multiple instances of Skanect with the newer versions. I haven't been able to launch any (either Mac or Windows).

Is there an argument or config file change that needs to be made in order to do so?

 

(Searching the group here I didn't find any references to it other than pre 1.5 versions.)

Thanks,

David

[Nicolas Burrus]

Hi David,

On Mac, you'll need to launch Skanect from the command line:
/Applications/Skanect.app/Contents/MacOS/skanect . If you enter this
twice in the console, it should run two instances.

On Windows, just launching the Skanect binary twice should work. What
happens when you do so? Do you get a message saying that only one
instance can be running?

[Tobby Ryan]

I too get this error when trying to launch 2 versions of Skanect 1.8.3

[eRoom-D]

For the Mac, I was running open -n /Applications/Skanect to make sure that a new instance would run, but it just went to the already opened instance.

As I was writing this I was tracking down to make sure I was running 1.8.3, and I opened the Scannect.app and looking at the Info.plist file there is a line for:

Application prohibits multiple instances

and it is set to YES.

Switching it to NO allows me to run multiple instances.

MacOS 10.10.4 Yosemite

Skanect 1.8.3 Pro

For Windows, it gave me the familiar "An instance of the application is already running". I haven't been able to try any methods of changing settings on the Windows machine, just launching the application from the Start menu.

Windows 7

Skanect 1.8.3 Pro

If you need anything more to help you debug, please let me know.

FYI, I'm a front end (Angular, HTML5) and mobile dev, and I am comfortable with the command line, so if you have any specific things to test I would be comfortable doing so.

Thanks,

David

[Nicolas Burrus]

Hi,

Thanks for the feedback. After a double check, indeed, multiple
instances are not possible on Windows by default. The following
command line switch is required: "--allow-multiple-instances" to
enable it.

On Mac, launching manually should work, or indeed modifying the plist.

We did not want to enable it by default since for most users having
multiple instances is just more confusing.

[Tobby Ryan]

Tomorrow I will be testing my new scanning rig. It is ready tonite, and has pre-tested as in wiring with stationary object works.

I will be opening up 2 versions of Skanect and letting everyone know how it performs

As a FYI, I have built my rig on  Fred Kahl's design.(Fredini) from thingiverse

The makerslide rig is basically per exact although I did cut my extrusions at wrong lengths, so its a little off. not much.

on the carriage plate I substituted the RocketFish Wall&Tv mounts. I used these as they were prebuilt. I took off the wallmount area and drilled 2 holes that matched the makerslide carriage then re-assembled the tv back mount. I put the m5 screws inside it as they are on the other side of the v-rollers.

Yes the top = top roller opposite

bottom = bottom roller opposite

[Tobby Ryan]

eh.  i tried. it switches back and forth every second between the two scanners when i hook them both up before opening.  does it capture data and streams from both, yep. but in the live data it looks bad. so i will go back to one for the 3 day scanning session at the end of this month.

i did however try connecting the second via a pcie usb 3.0 card, after the first one opened. i timed it right once to get drivers auto loaded and then the second instance of skanect.

on my gtx970, i was getting 30.97fps with one scanner,and around 26fps with both running at same time. core i7 3.05ghz processor.