Hi all,
i am using ipisoft for almost a year, but now i want to try developing some new features which ipisoft does not provide. to do that i need to use the data i already recorded and i have those in ipi file format. is there any way to get the raw data from the recordings? (ipivideo format)?
I am using kinect v2
thanks in advance

Hi
Currently we do not provide a specification for iPiVideo format, nor an API for working with it.
What exact data do you want to extract - color, depth, skeleton?
Also, can you share what features do you intend to implement?

I need only the depth data, i have some recordings of people sitting on the floor, and i just cant get any good resault in this situation so i want to try fitting a half body model to get the movments of the arms and torso.
thanks

Can you share one of such recordings for us to have a look?

I can't say at this moment, i have to talk to my superiors.
in any case i won't post it here, can you give me your email?

You can send information you don't want to be exposed in public to our tech support email, or PM me on this forum.
Upload an iPiVideo file to file sharing service and provide us a link to download. If you use multiple Kinects then please also send a corresponding iPiScene file or calibration project with camera positions.

I uploaded a short record (cut from the full one, normally my recordings are 40-60min).
this is the link
https://we.tl/Rnd0PloFmf
thanks, Ian

Thanks for sharing. I've taken a look on this record.
Indeed, it can hardly be tracked by Mocap Studio. The issues I see:

• Large distance between actors and sensors (about 4m). The quality of depth data provided by Kinect decreases with distance from the sensor. Obviously, the number of depth pixels per body is low at that distance.
• Actors mostly take poses where body parts are close to each other (and often to the floor as well). That is, depth pixels can hardly be distinguished to belong to a specific body part.
• Child's pixels are partially confused with background (floor) because the limbs are pretty thin.

I suppose that color cameras (PS Eyes) may be more appropriate for capturing in these conditions. Though I'm not completely sure about this. However, using color cameras imposes addition requirements on actor clothing (solid colors) and environment (bright even lighting).

Regarding using depth data from your recordings. I saw in your other post that you already own a license for Biomech add-on. Can't you just use point cloud exported via Biomech add-on as raw input for your purposes?

yeah i know that ps eye cameras will perform better but in my case it is impossible because i cant expect a collaboration from the "actor".
so my only option is to use the kinect setup.
did you looked at the calibration video, i suspect that my problems comes from there, because there is some deviation between the depth cameras but i cant make that better (i tryed over 100 time).
the licensed software is at work so i will have a look at the point could tomorrow, i am not familiar with this option.
to start work on a fix i need the full depth data from 4 cameras or the combined one (3D model).

the problem is the camera ' position too far away from the actor ,and get no better result not the ipi studio problems thanks