PDA

View Full Version : Did anyone use the kinect-one device with the last Nevronmotion version (Nov. 2015)



vipvip242
11-18-2015, 09:07 AM
Now Kinect V2 is apparently supported since november 2015 ( Nevronmotion v1.1):

-> did anyone try it ?
-> is it better the the V1 version in term of motion capture quality?
-> The kinect one device required is it the same as the xbox-one kinect or is it a 'special' windows version ?
-> Is it an USB adaptator required to plug it to a pC ? Which one ? compatible with Windows 10 ? Is there still anytrouble with spedific USB3 chipsets ?

So a bunch of question... any answer will be highly appreciated because it seems there is a few informations around that on newtek's website...

Thanks

TIMMYLYNN
11-18-2015, 06:05 PM
from what I've been reading Microsoft stopped making Kinect v2 for windows because of demand by smart people /devs and they can't fill the orders so they stopped making it and now only making Kinect for xbox one which is suppose to be same as v 2 was b4 they stopped making it which I understand supports up to six skeletal figures at once...please correct or add any info....so my question is from what i'm reading it plugs into xbox so how do you plug it into pc? I don't use xbox and don't plan to but I would like to have a cam to use with newteks software so if someone from newtek could tell us what device can be used to capture motion for this...is it only Kinect by Microsoft/ no generics yet??? what are the model numbers and what cables are needed? it says you have to have a usb 3 port for it to work....any input please...thanks in advance...

TIMMYLYNN
11-18-2015, 07:19 PM
delete

stevecullum
11-19-2015, 01:26 PM
There is an adapter from Microsoft that will allow you to plug the XB one device into a PC. I think it around 30.00

If anyone gives it a go, would be great to see the results. I tried Neveron with the original Kinect and the results were OK, but not great if I crossed limbs etc..

rkatos
11-19-2015, 05:25 PM
Hello, just beginning on this quest. Looking for Xbox One to Windows 7, found the following on the xbox.com website. Now, no information from me on whether this works with NevronMotion:

From xbox.com:
------
When you connect an Xbox One Wireless Controller to your Windows 8.1 or Windows 7 PC with a micro-USB cable, Windows will automatically download and install the drivers for the controller. If Windows is not set to automatically download device software, check for updates manually in the Device Manager:

Point to the lower-right corner of the screen, move the mouse pointer up, and then select Search.
Enter Device Manager in the search box, and then select Device Manager. Windows may ask you for an administrator password or to confirm your choice.



Locate and expand Microsoft Xbox One Controller.
Right-click the expanded Microsoft Xbox One Controller and select Update Driver Software.
Select Search Automatically for Updated Software.

Notes:

If the controller doesn't respond, try a different micro-USB cable.
To use your Xbox One controller with your console after using it with a PC, you must re-sync the controller to the console. You can do this by using the wireless sync button or a USB cable. For details, see Connect a wireless Xbox One controller to your console.
For Windows 10 users, the Xbox One Wired Headset and the Xbox One Stereo Headset are compatible with a PC when connected through the Xbox One controller, as long as the firmware has been updated to the latest version. To ensure you have the latest firmware for your controller and headset, see Update your Xbox One Controller.
---------

(( the url for this: http://support.xbox.com/en-US/xbox-on-windows/accessories/connect-xbox-one-controller-to-pc ))

Hope that helps. If I can get an xbox one and get it work, I'll post the result.

Marander
11-21-2015, 05:50 PM
@rkatos: your description is about the game controller, not the Kinect. You'll need the Kinect for Windows Adapter ($49) to connect the Kinect to the PC.

vipvip242
11-21-2015, 11:25 PM
OK, i've ordered a kinect-one + adapter (and perhaps 2 kinect-one in a second time if the first test is decisive), and will report to this post after trying it...
Crossing the fingers all will work as expected because with the kinect-360 (not recommended by newtek, they recommend Kinect1-for-windows device), it works but it is crappy...
cheers

Greenlaw
11-22-2015, 12:01 AM
Kinect One (aka Kinect v2) is not supported by Windows 7. You will need to upgrade your OS to Window 8.1 or higher. Luckily the Windows 10 upgrade is free until next July.

You will also need a USB 3.0 controller as v2 is not compatible with USB 2.0. In my experience, if you have to add an USB 3.0 controller card, you'll need one that lets you tap into your computer's power supply. The motherboard should be a fairly recent model too because older motherboards may not have the necessary throughput for full USB 3.0 support.

Additionally, you will need a graphics card that supports DirectX 11 or higher.

Microsoft has a Kinect Configuration Verifier tool which will tell you if your computer is Kinect v2-ready. You can download this software from their website.

If your computer meets the above and is a fairly recent build, it should work fine. I have a Wacom Cintiq Companion 2 tablet computer and it's been running fine with Kinect v2.

G.

P.S., I have some information about using Kinect v1 and v2 for motion capture in the Brudders Production Log (http://forums.newtek.com/showthread.php?133274-The-Brudders-2-Production-Log-%28Well-sort-of-%29). Not Nevronmotion related but it may offer some insight on hardware/software requirements.

Greenlaw
11-22-2015, 12:21 AM
One more thing: if you need to extend the cable, be sure to get an Active USB 3.0 cable (aka, USB 3.0 Repeater cable.) This type of cable can boost the power/signal to maintain an steady data rate for devices connected to it. If you use a regular USB extension cable, the Kinect One (Kinect v2) will probably fail.

vipvip242
11-22-2015, 12:46 AM
Thanks for these infos : indeed i've already read these contraints on other forum and my system should be ready for this ( win10, DX11 card and specific USB3.0 chipset : only renesas or intel ones apparently...).
regards

***update*** : i' just launched the kinect-config tester and the system seems to be ok, exept the kinect-V2 adapter missing, of course... thanks for this tip, very usefull !

Greenlaw
11-22-2015, 01:15 AM
Actually the info you posted may possibly have helped me. Earlier I had tried to make a Kinect v2 work with one of my older workstations but it failed. I'm wondering if it's because the USB 3.0 card in is has the NEC chipset, not Renasas. Fortunately, we have a few spare Renesas cards here I can try out.

Thanks for the tip--this info, if it works, may allow me to try a triple-Kinect v2 configuration.

Greenlaw
11-22-2015, 01:21 AM
Oh, wait...Renesas is the NEC chipset. Oh, well, nevermind. I guess I should be satisfied with dual-Kinect v2 for now. :)

vipvip242
11-22-2015, 01:33 AM
Please give us feedback too if your configuration is working : i'm curious to know if 2 ( or 3 ) kinect V2 greatly improves the mocap ( i've read 2 is necessary to get full 3D front/back movements ) : i'm looking for a shematic explantion showing the optimised positions/distances of 2/3 adapters .
I've some free surface in front of the green screen studio, but it is not hollywood ;).
If anyone find these infos, thanks to share !

Greenlaw
11-22-2015, 01:50 AM
FYI, you cannot use mutliple Kinect devices (v1 or v2) with Nevronmotion because it uses the Microsoft Kinect SDK. The official SDK only allows you to use one device per computer.

For multi-Kinect recording, I've been using a third party program called iPi Mocap Studio, which allows you to use up to three Kinect v1 devices on a single computer or up to four Kinect v2 devices distributed across four computers. As far as I know, iPi Mocap Studio is the only program that currently does this.

FWIW, the software works great for this--smooth motions, good feet tracking (compared to single Kinect), and relatively few errors for a wide range of motions. During 2013-2014, I used the dual- and triple-Kinect v1 system along with multi-PS Moves for head and hand tracking for digital stunt double and creature effects in a few feature films. Of course, I also use this system for my own animation shorts--my first use of multi-Kinect was back in 2011 when I made 'Happy Box', which was the first animated short created anywhere using multi-Kinect capture.

Lately, we've been experimenting with multi-Kinect v2. The main difference with Kinect v2 is that the higher resolution results in reduced noise and the wider FOV allows for a larger capture space in the same size room. The framerate is still 30 fps, so it's not ideal for very fast motions. Fortunately, Mocap Studio does well at compensating for the low framerate. Using multiple devices also helps minimize errors caused by fast motion.

I had posted more detailed info about Kinect v1 vs v2 and Nevronmotion in the LightWave blog thread last week. There's also more general info about his in the Brudders thread mentioned above.

Hope this helps.

G.

vipvip242
11-22-2015, 02:02 AM
thanks a lot for these precious informations !

Greenlaw
11-22-2015, 02:07 AM
Here's a very old example of a dual-Kinect v1 test recorded back in 2011 when my daughter was five years old: Mocap with Small Child (https://www.youtube.com/watch?v=PJZabVZqCl0). The dual-Kinect capability was an experimental feature at the time and the developers were actually surprised that we were able to make it work with such a small subject--this version of their program was only meant to work with full grown adult bodies.

I think it worked remarkably well for early beta software that had never been tested this way before. As a matter fact, in the following release the developers responded by making changes in the code for us to improve motion capture with children. :)

The software has seen many advances since then--tracking is smoother and more accurate (we now have head and hand tracking) and the editing tools have been greatly improved.

vipvip242
11-22-2015, 02:15 AM
that's very clean ! the problem with Ipi is the price of the-per year licence... i've seen brekel, much cheaper and without rental licence... did anyone experienced it ?
Thanks

vipvip242
11-24-2015, 12:48 PM
I've just received the kinect-one sensor and his adaptator : big deception, it doesn't work :
the kinect sensor is recognised, but nothing happens when lauching studio ( active/live enabled with the kinect-one skeleton ). The same procedure worked with the Kinect 360 adapter...
On the system (win10), the sensor is well working either with the SDK 2.0 tools and with the tryout version of Bekel bodycapture V2, which uses the SDK too...
Any help would be highly appreciated... or is it a bug of the plugin ?

Greenlaw
11-24-2015, 01:02 PM
I've only recently started using v2 sensors with iPi Mocap Studio but I haven't tried a v2 with Nevron Motion yet.

From what I recall, some users reported needed to copy a certain .dll to Lightwave to make it work with v1. I don't think I had to do that (with v1) but, like I said that was some time ago. That said, I'm not sure this applies to v2.

Since I have everything set up right now, I'll try the v2 with Nevron Motion tonight and let you know how it goes.

G.

jasonwestmas
11-24-2015, 06:01 PM
I've only recently started using v2 sensors with iPi Mocap Studio but I haven't tried a v2 with Nevron Motion yet.

From what I recall, some users reported needed to copy a certain .dll to Lightwave to make it work with v1. I don't think I had to do that (with v1) but, like I said that was some time ago. That said, I'm not sure this applies to v2.

Since I have everything set up right now, I'll try the v2 with Nevron Motion tonight and let you know how it goes.

G.

So lightwave 2015.3 sees the kinect one (xbox one kinect) in the VS device manager but it has no settings button (it just says "none" where the word "open" should be but the device is also "ready" and the kinect is in fact lighting up when I check mark the enable button) which should allow one to open a new window to see what the camera sees. Any ideas on that? I'm pretty sure I installed everything that needs to be installed. Windows 10 says the kinects is working properly too so, not sure where that window is.

vipvip242
11-24-2015, 09:21 PM
Thanks guys, for following it...
Jason : yes, you're right, it is strange : no way to open the camera in studio settings, like it was the case with V1 version and what is shown in NT tutorial " Getting start with Nevronmotion & Kinect"....
Perhaps this version is simply not working, it would be cool if someone of Newtek could inform us about this...

Greenlaw
11-26-2015, 10:47 AM
Sorry for the cross-posting but I got it to work.

131208

Here's my daughter waving. Here's my daughter waving. I captured this on my tablet computer running Windows 10.

It's actually pretty to easy to activate but, as mentioned in another thread, the docs make it more confusing than necessary. I'll post a quick walkthrough with a demo clip after Thanksgiving dinner today. 'Dinner' is going to be early so maybe this afternoon.

Anyway, yes it works and it definitely works better than v1.

G.

Greenlaw
11-26-2015, 08:01 PM
Here's a Kinect One/NevronMotion setup tutorial I whipped up this afternoon:

Setting up Kinect One for NevronMotion in LightWave 2015 (http://forums.newtek.com/showthread.php?148971-Setting-up-Kinect-One-for-NevronMotion-in-LightWave-2015&p=1457058#post1457058)

I hope it's helpful.

G.

vipvip242
11-27-2015, 04:38 AM
So i followed your instructions : it works : many thanks !
Apparently the problems, on my side, were:
* the kinect-one device was not renamed
* the HID was not enabled

stevecullum
12-02-2015, 06:38 AM
Here's a Kinect One/NevronMotion setup tutorial I whipped up this afternoon:

Setting up Kinect One for NevronMotion in LightWave 2015 (http://forums.newtek.com/showthread.php?148971-Setting-up-Kinect-One-for-NevronMotion-in-LightWave-2015&p=1457058#post1457058)

I hope it's helpful.

G.

Thanks for posting this. Are you able to capture motions where limbs cross over each other? Lets say you were wielding an Axe above your head and then bringing it down in front of you. Or maybe standing up from a crouched position? (Like the Terminator entrance!). These types of movement caused plenty of issues for V1 and wondered if V2 improves things...

jasonwestmas
12-02-2015, 07:30 AM
Thanks for posting this. Are you able to capture motions where limbs cross over each other? Lets say you were wielding an Axe above your head and then bringing it down in front of you. Or maybe standing up from a crouched position? (Like the Terminator entrance!). These types of movement caused plenty of issues for V1 and wondered if V2 improves things...

No you still get the occlusion problem. You'll still need something like IPI for most complex poses because Lightwave doesn't support merging the data from two kinects together. The hands and feet suffer the most from this, even when standing in the center of the world of the camera like you are supposed to.


Outside of that I was very impressed with how fluid the motion was in real time with a single kinect one. The only thing that stood out the most (during simple poses) is how the spine is so rigid during recording and simplistic. That can be fixed by hand if your final rig that you are retargeting to has enough spine joints. But again that defeats the purpose kinda.

stevecullum
12-02-2015, 09:39 AM
Ah ok - I guess you can't expect a home baked single Kinect solution to replace a pro system! Still a very useful plugin for transferring motion data to a character though.

jasonwestmas
12-02-2015, 10:56 AM
Ah ok - I guess you can't expect a home baked single Kinect solution to replace a pro system! Still a very useful plugin for transferring motion data to a character though.

yeah, the Nevron Motion package is really two things, the retargeting plugin which is nice to have in your native environment to save some time. Then there is the Visual Studio addon which works with the kinect to drive motions to any object in the scene really, like a puppet, using your body (and all with nodal power at your disposal) thanks to the Visual Studio part of lightwave.

Greenlaw
12-02-2015, 10:57 AM
Are you able to capture motions where limbs cross over each other? Lets say you were wielding an Axe above your head and then bringing it down in front of you. Or maybe standing up from a crouched position? (Like the Terminator entrance!). These types of movement caused plenty of issues for V1 and wondered if V2 improves things...

I haven't really used a single sensor since we made the original 'Happy Box' years ago, but I did find that you can get some complex poses with a single sensor by meeting it halfway.

For example, perform the pose close to what you need but keep your limbs apart just far enough so the device can register their positions. Then, in Lightwave, you should be able to offset the rotation of those limbs to complete the pose. For example, in the axe swing, keep your hands apart during capture and afterwards you can bring them together by offsetting the shoulder rotations. For the crouch pose, crouch about halfway and then offset in post to push the crouch further. (You'll probably want to design your rig with rotation offset bones built in, or use IK Boost. That or use a third party tool that allows you to layer your animation offsets.)

You can also fake some complex motions through clever editing. For example, the original version of the 'chainsaw dance' in Happy Box was recorded using a single Kinect and the performer was standing in place and only acting like he was spinning around. In LightWave, the entire rig was simply parented to a null and rotated for the 360 degree spinning motion. This motion was eventually replaced with a dual Kinect capture with real spinning motion but to be honest, there was actually very little difference between the two versions.

Another tip: try to visualize what the sensor is seeing when you perform. Remember, it can't track what it can't see but for some motions, you can help it see more by adjusting your orientation from the sensor. The sensor may be a 3D recorder but it still needs to see a limb to know how far away it is in Z-space, so make sure it can see everything it needs to to capture your pose. For example, it might be easier to record that 'terminator' crouch pose if you start out standing in a 3/4 view, making sure all limbs are at least mostly visible during the motion.

Regarding iPi Mocap Studio and mutli-Kinect, yes, these kind of motions can be captured much more accurately with multiple sensors, but even then you sometimes need to adjust your performance for the system and then tweak the motion in your animation program to get the motion really want.

In other words, don't fight the system by making performances that it can't possibly capture--that's likely to generate a lot of errors which can be very tedious to fix. Instead, adapt your performance to what the system can actually record and then enhance that performance in animation and editing. IMO, enhancing animation is usually easier and more fun than fixing a whole bunch of errors.

Hope this helps.

G.

Edit: Regarding v1 vs v2 for these kind of poses: I believe v2 can do a better job because of the improved SDK. v2 records in higher definitiion but this mostly helps eliminate or greatly reduce the jitters more than it improves capture accuracy. What would REALLY improve accuracy is a higher frame rate but unfortunately Kinect One (v2) is still records only 30 fps.

stevecullum
12-02-2015, 11:37 AM
The compromised approach seems like it could work. Hadn't really considered mixing with post capture adjustments. I'll have to see if I can find someone that will loan me a Kinect V2 to experiment with! :)

Greenlaw
12-02-2015, 11:52 AM
You can also use IK and Constraints to override parts of a peformance to improve its accuracy.

For example, in B2, when Sergeant is playing his harmonica, I pantomimed this motion by imagining a very large harmonica. The recorded motion was close to what I wanted but the mouth and paws were not really aligned with the CG harmonica so I used constraints on the head and paws to make the performance look more correct. The mocapped data is still there in the final animation but it's been enhanced with IK for accuracy. (I used Motion Builder to edit this performance but you should be able to do it in LightWave too.)

stevecullum
12-02-2015, 12:15 PM
You clearly have plenty of MoCap tricks up the sleeve! I have some iffy mocap walk cycles that need the feet sticking to the floor - I guess the IK trick will work for something like that too. You should do a commercial 'Working with MoCap in Lightwave' video - I'm sure there a plenty of folks who need that knowledge now Neveron was given away.

Greenlaw
12-02-2015, 12:42 PM
Thanks! Yes, like Felix the Cat, I have a bag of tricks for almost any occasion. :D

I keep threatening to make more tutorials but I still have two big personal projects to complete. When they're done, more tutorials and maybe even full training courses is going to be my next goal.

G.