Hi,

How do the Primsense Carmine sensors - now with support in iPi Recorder - compare to Kinect sensors? Do they improve the motion capture in some way?

Hi

PrimeSense Carmine is essentially the same device as ASUS Xtion Live released under a differend brand name - the name of the developer of underlying depth sensing technology. So everything that has been said in comparison of Kinect vs Xtion can be applied to Kinect vs Carmine.
http://wiki.ipisoft.com/Depth_Sensors_Comparison

These guys seem to say the Carmine provides better tracking results.

any reason why this could be the case? Or you think perhaps just their opinion ;)

http://www.faceshift.com/news/carmine-1-09/

The Carmine 1.09 works at close range (I think about 30 cm) at 60 fps, where as the Kinect natively runs at 30 fps (from what I remember). This works well for Faceshift as the face has much more smaller nuances than a full body has. The extra frames serve to pick up on these motions while the closer range helps to deliver a clearer image quality. It would work great for faces but not necessarily for full body mocap. The Carmine 1.08 would be the full body equivilent but the Faceshift guys are more interested in the 1.09 model. I could be wrong though, I use PSEyes :P

Xtion does not have a separate "near" mode. Kinect produces more noisy data than both Xtion and Carmine. From this prospect, "near" mode Carmine 1.09 may be the best choice of three models for face capture.
In our applications, we do not see any considerable difference between Xtion and Carmine 1.08 (which is a "far" mode model). 60fps modes of Carmine are still too noisy for capturing, as it is with Xtion - it seems to be the hardware limitation of the current implementation of depth sensor, which is the same in both models.