For what it's worth, I avoid the 180 degree setup because I seem to get far more trackable results when using the 90 degree setup. To me, the 180 degree setup appears more susceptible to IR interference and could be less accurate. My primary test case was the two '3 Happy Cats' sessions--the 90 degree session tracked fine with only a few difficulties involving collision (described elsewhere) but the 180 session was almost completely untrackable. Granted, the motions my daughter made for each session were not exactly the same but I think they were similar enough to be a fair test.
Other factors can vary quality as well, such as the environment and the type of motions being tracked but between 90 vs. 180 degree setups, I really am getting better results with 90, at least in the few comparison tests I've done. My guess is that the volume overlap that exists with 90 degrees gives Studio a more accurate shape to work with; when you record with full 180 degrees the volume overlap does not exist at all. Also, with 180, the signal strength probably isn't as constant as when you record with 90--for example, with 180, when you walk towards one device, the signal may get stronger but it proportionately gets weaker with the other device; with 90 degrees the strength between the two devices may stay a little more balanced as you move between them. That's just a guess though--be advised that these are unscientific observations and you should probably take my ideas about this topic with a grain of salt.
As for why the left hand seems more trackable than the right in your example, I don't know the answer--I haven't seen that problem here and I can't even guess what may cause that. Note: to date, I've only recorded data using iPi DMC 1.0. Just wondering, could this be a new Studio 2.0 problem? Or was your example tracked using Studio 1.0? I'll try to do some controlled tests tomorrow using 2.0 and Kinect and let you know if I notice anything like this.
(Yay! I finally have a day off so I can spend it working.) :)
Just curious but when you make your T-Pose, do you directly face a Kinect or do you have your arms pointed towards each Kinect? When I did my 180 degree test, I faced one device directly. When I do my 90 degree recordings, I face the left side Kinect, though I don't think it really matters which one you face. I think I've also tried facing the space between the Kinects, though I can't remember if that was better.
Back when we did 'Happy Box', I believe I faced the space between the two Kinects and the devices were only about 60 degrees apart--I haven't compared to see if this was any better as I assumed a full 90 degree would be better. That said, at 60 degrees there is even more volume overlap which might improve accuracy but only when facing the devices--obviously, there is increased risk of occlusion errors as you narrow the range of visibility.
Note: I just checked
this video from last year and it does look like I faced the space between the two devices. It also looks like my range was even narrower than 60 degrees. Surprisingly, this configuration worked fine even for the 'chainsaw dance' shot where 'Sister' spins around several times. I don't recommend going less that 60 degree though--better to go full 90 if possible.
BTW, it may seem to make sense that you would have less chance of occlusion errors using the 180 degree setup but because you're recording depth, occlusion isn't nearly as problematic with 90 degrees as you might think. When recording depth from 90 degrees (as opposed to just 2D video), you're getting nearly complete 360 coverage--you almost have to make a conscious effort to hide things from both Kinects to produce occlusion problems.
Anyway, this is just my personal experience and opinion. Hope some of this is helpful but please continue to report your experience as I'm curious to learn why we're getting different results.
G.
