Tagged: 17
- This topic has 18 replies, 4 voices, and was last updated 7 years ago by casetero.
-
AuthorPosts
-
2017-08-27 at 12:51 #298caseteroParticipant
Hi there, i have been testing the kinect v2 with driver4vr. Im using it with PIMAX 4K.
The thing is i just want head tracking for seated games (asseto, il-2,…). I cant get the kinect tracks my head correctly, in the settings of driver4vr the seated mode doesnt work, and it tracks my legs and my arms in a totally weird mode so i guess it breaks the track of the head. When i put the PIMAX in my head is even worse because the kinect detects the glasses as another arm or something.
Is there a way, via windows kinect SDK 2.0 or something, to say to driver4vr that just tracks the head? I think the solution maybe easy if the kinect would track only the head and not the lower part of the body and the arms.
Thank you very much.
2017-08-27 at 17:24 #299JazneoParticipantthat Sept 17 update he didnt release it yet
2017-08-27 at 18:24 #300caseteroParticipantWell, that´s just the opposite. 🙂 But it seems great anyway!! I refered to add a seated option to he kinect v2, and not tracking the lower body and arms.
I´ve been trying around with te kinect v2 and ive reached some decents results, but not great. I have the kinect positioned up the monitor and i play seated games. So i play close to the kinect (700 mm), that makes some weirds movements in the Y axis when moving the Z axis. This movements are rapids, like jitter. I think the problems come because driver4vr take the data directly from kinect SDK 2.0 and output it. If you could make some kind of filter it would give us a smother experience.
Thank you!!
2017-08-28 at 01:10 #301Du1g0ParticipantMmmmm….i think it could be just a case of you being too close to the Kinect…that can cause the problems you describe….
2017-08-28 at 07:32 #305caseteroParticipantI did more experiments with the position and the angle of the kinect relative with the PIMAX, got more accurate movements. If we had any kind of SDK prepared to work with low distances (or seated tracking) and heads with virtual HMD´s i think it would work great.
HAve you seen this? http://www.immotionar.com/
Another thing we need is some kind of filter to make smoother the experience while playing. Maybe im not introducing the values correctly in the settings of kinnect. Any kind of guide to this settings window would be great too.
Thank you!!
2017-08-29 at 07:20 #308Greg DriverKeymasterWith current Kinect I am using skeleton tracking and I rely mostly on what Kinect provides in skeleton tracking…
I created topic for new feature to improve headset tracking quality.
Let me know if any solution would be ok for you. If there is anything I need clarify let me know.
Greg2017-08-29 at 08:03 #314caseteroParticipantHi there Greg.
I think it is better not having any external object with kinect, if i understand correctly for this purpose you would use the normal webcam of the kinect, aren´t you? If this is true we are gonna loose the depth propierties of the kinect, and i think is quite usable this depth camera.
Yesterday i was trying things and i did a configuration that looks pretty promising, but i need more test in it. The string si the following:
Opentrackkinect -> Opentrack -> Freepie -> Driver4vr —–>>>>> SteamVr
Doing this configuration you can benefeit of the Accela Filter in opentrack so the tracking is really smooth and you can scale it. Just tested it yesterday but looked promising.
I will update this post tomorro with more testing.
Thank you for all Greg!!
2017-09-01 at 07:30 #326caseteroParticipantHello again,
I´ve been doing more testing in kinect. I continue thinking that the big problem for the kinect is that it wants to track the full body (skeleton API), and that makes jitters and erratic changes of position when you are seated. The fact that you have a HMD in your head doesnt help too, because the kinect thinks that your body starts just below of your HMD and does erratic movements too.
This could be adressed making a new SDK prepared to track faces with HMD´s and with the option to track seated people (like kinect v1 has).
This problem improves with distance, if you can put the kinect in front of you, with an angle near to 0 and with 1.5 m of distance, the track improves. But it continues making some erratic movements, less often but it does. Also, for placing the kinect as i said, you have to do a new space to use vr that is not in front of your monitor, so i dont know if it is a good solution.
Another problem is the lag. The kinect itself introduces some lag in the ecuation, and the fact it works at 30 Hz, make it some laggy. Dont expect to be as responsive as rotational track (in PIMAX or GEAR VR). You can live with it, but it present and rest inmersion.
For the opentrackkinect vs native kinect v2 driver4vr option, i much prefer the opentrackkinect because of the smothness of Accela filter,i know this introduces some lag to the ecuation too but i need that smoothness to play. Furthermore the opentrackkinect take the input of the skeleton API too so it has the same problems and does the same erratic movements that i mentioned.
Hope this helps.
2017-09-01 at 10:01 #330Greg DriverKeymasterCasetero,
Can you increase Head position smooth in Simple physics section of Kinect ?
Also recuce Head speed factor. That should give you more precise tracking. Let me know.2017-09-01 at 10:54 #331caseteroParticipantHi Greg,
Actually i have dissassembled my kinect system and trying things with psmoveservice stuff. I will be on holiday all next week so i cant really test it again for at least 2 weeks.
A thing ive noticed is that i dont need freepie to make the conection between psmoveservice and driver4vr (just to make the head tracking, i dont use controllers). Is this normal??
Another thing id like is to have a smoother track with psmoverservice+driver4vr. Is there anyway to make this string works:
psmoverservice -> PSMoveFreePieBridge -> Opentrack -> Freepie -> Driver4vr
???
I think with this string i could reach the smoothness i want.
Thank you!!
2017-09-01 at 11:12 #332caseteroParticipantI forgot to say, that this smoothness is achieved in the tracker test of psmoveservice (the hmd moves smooth like butter in the test), but in steamvr it doesnt run so smooth, its like in the psmoveservice test i reach the 60 Hz and in steamvr im going at 20-30 Hz.
It is a big diference between the rotational track (given by PIMAX), and the positional track (given by psmoveservice + psmoveservicefreepiebrigde + driver4vr)
Thank you!!
2017-09-01 at 20:11 #338Greg DriverKeymasterI am afraid the problem is that I am smoothing Kinect position. You can see it in Kinect settings. If you parameter to 1 it will give you almost immediately position of head but it will be skaky. I must check if I can make any better way to get head position out of the Kinect and have some ideas. But I need release body tracking first as it is almost done.
2017-09-18 at 10:36 #457caseteroParticipantHello again Greg,
I´ve been testing the new version 2.0 of drivervr with kinect with its own driver for kinect, no opentrack.
I dont know if you have improved something with the positional track of the head, but now it works better for me. I can put the kinect at 45º degrees now (to the side) and it works. I would say it at least as playable as psmoveservice with a lot less previous work to do, but i have to do more deep testing.
Can you explain one for one the settings of the kinnect of driver4vr? I have been able to configure it, but i dont really know what do the differents parameters to configure.
Thank you for your work!!
2017-09-18 at 11:20 #458Greg DriverKeymasterhello. yes. i changed default parameters and reset them.
in general smoothness is how you smooth raw data. this is not best approach because in raw data there are samples that should be simply removed but now they are averaged. anyway this is simplest way to smooth position. if you want raw data as kinect gives then set smoothness to 0.second parameter is velocity or speed I think and it is how we get forecast based on current speed. so smoothness is lagging but speed is trying to guess where it should go next and in some way compensate lag. if it is too big then it will take too much of speed factor and make your head dancing.
then last parameter tells how many samples we take to calculate speed. the longer the biggest lag.
again. those are to play with but first need make initial filter of data from sensor that will improve for shaking and we can take less smoothing.
can u also place kinect on the back to your game orientation and set parameter rotation to 180. perhaps it will work better.
2017-09-18 at 20:57 #464caseteroParticipantHi Greg,
Thank you for the explanation!
With this knowledge ive been able to reduce the lag a bit, with the same smoothness. I still notice that raw data that should be removed. The main problem is that kinect, when facing the hmd to it, read the hmd position as if it would be my shoulders. And it blinks in the Y axis as result. My head moves quicly up and down when facing the kinect (just in some positions). So i guess that any kind of filter could be do it to remove this movements. Do you plan to add this kind of filter to driver4vr?
I have tested the back postion of the kinect to my game orientation.It works nice, but when you look down and move your head down too, the kinect loose the track of your head because it doesnt see it, and it doesnt track your down movement (i play seated).
Thank you!
2017-09-18 at 21:48 #467Greg DriverKeymasteryes. that kind of fake movement I hope to remove by filter.
but i also work on color led tracking. it does not need to be ball, perhaps line or small leds or something else. that should be reliable… i hope 🙂2017-09-19 at 07:40 #473caseteroParticipantOk.
I think if you code a good filter, the smoothness and reability of the depth camera of the kinect will be enough.
I already have a light ping pong ball for psmoveservice, but it needs another cable linked to my head and its work depends on the light conditions of the room.
I would much prefer a configuration with only kinect working, and if now it almost work with just averaging the raw data, i believe that when this raw data will be removed it will work nice.
Thank you!!
2017-09-20 at 07:52 #487Greg DriverKeymasterthere are really cool led light on small batteries that last for weeks. i don’t need bulb but only a source of light. I will show more details when we will get there.
but yes. will try first with code improvement.2017-09-20 at 10:18 #488caseteroParticipantOk,
Im sure you will provide something good to us!!
Thank you for your work!
-
AuthorPosts
- You must be logged in to reply to this topic.