Hey there,
I am wondering if it’s possible to record a user movement / contour / depth image from Kinect1 data stream somehow?!
Microsoft Kinect 2.0 Tools | vvvv seems to give huge file sizes (2gb for 14secs?!). I don’t really need a super realistic recording with background etc, only the contour maybe with depth.
In the end I’d like to view live user data and a recorded image (aka ghost) at the same time
Any tips on how to approach that?
Hi @yochee!
You may want to store the sequence dds files with Writer (DX11.Texture 2d), the file will be even smaller if you store it as a single channel texture, use ChangeFormat (DX11.Texture 2d) node to convert the Player (Kinect Microsoft) from B8G8R8A8_UNorm to R8_UNorm.
Just a side note, I never really liked the quality of Player texture for the purpose of masking and/or user detection. There are lot of errors both A and B type errors - part of user not marked as one or part of enviroment marked as user.
You will be better of by masking with bounding box and world texture.
hey yochee,
just to add my experience on id144’s answer, I used elliot wood’s ReadbackDX11 to record kinect streams as image sequences many times without any problem.
(@id144 thanks for the boundingbox/world texture tip by the way!)
hey guys, thanks for answering.
for now I tried // kinect-recoding-02.zip // in
even though writing to ssd (i7-4720hq, gtx970m, 16gb ram laptop) the process slows down or blocks the actual live view of the kinect in the renderer.
For now I will try with @id144’s advices and hope to get more performance. Otherwise I will try the readback thingy, which approaches parallel vs async reading/writing :)
hey id144, once again thanks for your tips. the writer does its job and it is a lot faster than the “Emulating Kinect data” above. ReadbackDX11 seemed to complicated for me
The quality of the Player texture with Kinectv1 is really bad, indeed. But it is small when recording and I just add some DX11.textureFX to make it look better. Best so far are:
Median, Blur, NormalGlow(!)
(…still investigating)
1 ChangeFormat really saves 20kb per PNG, but the colors are a bit different (quite stripped)
2 I am playing with your kinect hitboxes! Do you think its easily possible use it with kinectv2, I didn’t try the v2 yet. It might help with the quality (but not too much I guess)
3 I didnt get the solution with bounding box and world as I don’t know where to connect boundingbox
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.