I have had some good results with the very latest revision, but I am still plagued by a troubling amount of yellow noise both in calibration and on the actor doing mocap, even after putting up cloth backdrops to eliminate noise in the background. I also seem to be ending up with a capture volume smaller than expected based on the documentation. I want to make sure I am getting the basics correct, so my questions are these:

Assuming an actor who is 1.9 meters tall, what would be the optimal distance from the actor to each of the two kinects, and what would be the optimal height off the floor of each of the kinects? By "optimal distance", I mean the distance that would maximize the capture volume without loss of resolution.

Also, a side question: if an actor is taller, the area within the capture volume where his entire body will be visible is less -- therefore, is it correct to say that a taller actor therefore means an effectively smaller capture volume?

Hi,

Yellow background noise shouldn't be too much of a problem. What you want to watch for is too much seeing yellow noise on the floor because that can affect how iPi Studio finds the gound plane. My understanding is that is Studio can establish the ground plane and just enough bg info to establish distance, it should be okay--the details in the far background really shouldn't matter since you' probably won't be interacting with it.


Are the Kinects placed closer to 90 or 180 degrees? My experience is that the 180 degree setup can be more accurate for some motions but I think I can get a larger space using a 90 degree setup. Typically, I find that the quality of the 90 setup is just fine and may better for motions where you're generally facing the cameras. FWIW, I've successfully tracked 360 motions just fine with 90 as well. (Some online examples of 90 degree capture of 360 motions 'the chainsaw dance' in Happy Box and my daughter's 3 Happy Cats motion test.)

In general, the size of the capture space can be as large as the region in which you can fully see the actor in the depth data in both sensors. If the Kinects are optimally positioned, this can be around 7 x 7 feet, but keep in mind that the data quality is less accurate the farther away the performer moves from the camera.

Tip: when setting up the cameras, place a chair in the middle of the space or a book on the floor--or even better have a person stand there. Try to line up the cameras so that the item or person shares the same space to both cameras. This will help will defining a optimal region that is shared by both Kinects. Once the cameras are aligned, mark the center point with some tape--this mark will give you a consistent position for calibration and T-Poses. Also, mark the visible boundaries for the performer--when doing this, make sure the actor is visible from feet to head, with some clearance above for raised arms. If you don't anticipate the need for raised arms when the performer is near the camera, you may be able to expand you capture space a bit more.


Yes, that is correct. As mentioned previously, you can make the space a little bigger if the performer does not raise his arms above his head or jump up high for that matter when he is near the cameras.

Hope this info is helpful. Once you've done a few sessions, you'll get a feel for what the limits with a given setup and figure how to adapt and work around the system. Eventually, you will learn that some setups are more optimal for specific motions.

Good luck and have fun!

G

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!