Hi! We actually have a Unity project that can be used to display the point cloud from LiveScan3D in VR. You can find it here: github.com/MarekKowalski/LiveScan3D-Hololens You can also playback previously recorded sequences using the LiveScanPlayer tool available in the original repo, here: github.com/MarekKowalski/LiveScan3D

+James Sheridan Hi James, The point clouds are aligned automatically using visual markers. You print the marker on a piece of paper and attach it to something rigid. The Kinects see the marker and infer their relative poses from it. We also plan to include calibration that will use the skeletons detected by the Kinect For Windows v2 SDK. When this feature will be included, no markers will be necessary for calibration. Marek

+Marek Kowalski your just using the checkerboard detection in opencv? is there any documentation/instructions for this or if I try and run things or look through the code it should be obvious?

+James Sheridan Hi, we use our own custom marker system. We do not use the checkerboard, because we want to be able to distinguish between different markers that are present in the scene. There is a manual for the software included in the files, and starting everything should be straightforward. Some settings are still not included in the manual, but we will supplement it soon. If you want a more technical description, there is an article from 3DV 2015 that should be published soon. It is titled: "LiveScan3D: A Fast and Inexpensive 3D Data Acquisition System for Multiple Kinect v2 Sensors"

+Marek Kowalski - Thanks a heap Marek and top work

+James Sheridan Thanks a lot James.

+Ghassem Tofighi Hi Ghassem, thanks for the comment, as for your questions: 1) The positions of the cameras are different in each of the sequences in the video. In the first sequence there are three Kinects located very close to each other but facing different directions (outward facing config). In the other two sequences the Kinects are located around the object at around 90 degree angles (inward facing config). 2) If you use the outward facing config, you will need 6, since each Kinect has a horizontal FOV of 70 degrees. Using 4 Kinects and placing one in each corner of the room seems more reasonable. 3) In our system, each Kinect is connected to a separate computer, so in terms of processing power, including more Kinects is fine as long as you have more computers. The thing I would worry about when adding more Kinects is interference, the more of them you have looking at the same scene the more likely it is you get interference. This does not happen as much in the outward facing config, since the overlap between the devices is small.

@@MarekKowalski1 Cool project! Maybe you know this approach towards interference? www.precisionmicrodrives.com/content/using-vibration-motors-with-microsoft-kinect/

+Brian Handy Hi Brian, thanks a lot! Getting the app to work with libfreenect2 would indeed be very interesting. It would simplify the setup (less computers required) and probably increase the framerate as well. Moreover modifying my app for libfreenect should be very easy, it would only require implementing a certain interface class for libfreenect (it is already implemented for the Windows driver). In the last several months I have been very busy with other projects so I didn't have time to implement new stuff for this app and that probably won't change in the near future. If you are interested in trying to get libfreenect to work with LiveScan3D, I would be more than happy to help you!

+Marek Kowalski I think I actually already got my own code in libfreenect2 working. :) But your code is much more fleshed out, so I may switch over! donotunplug.tumblr.com/post/143541265027/kinect-videogrammetry-in-motion I have two questions for you even so though! I'm curious either way, 1) if you store the data to disk (and how), as I found no videogrammetry file formats online, and 2) how you do your alignments. A quick glance at the repo makes it look like iterative closest point, but is that a custom implementation or from a library like PCL? Thanks! And I may take you up on that offer yet!

+Brian Handy Hi Brian, I'm glad you like the app, hope that you find it useful :) As for your questions: 1) Each recorded point cloud (each frame) is saved to a separate .ply file. As you probably know .ply is a format used mainly for meshes and point clouds and not really a 3D video format. To be honest I don't know any 3D video formats either. 2) The camera poses that are used for alignment are obtained using visual markers. A marker is a pattern printed on a piece of paper and attached to a rigid surface. ICP is an optional step that can sometimes be used to refine the camera poses obtained using the markers. Having the option to use libfreenect as well as the official SDK in my app would be a great feature. However if you only want to use a part of my app for your own purposes (for example the marker detection and pose estimation), I believe the license allows it as well. :) Cheers! Marek