Pose model (hands recognition) - Multiple bodies - What's the proper color?

[3dcinetv]

Hello, I'm getting bad results all around when I try to push my own image into the POSE model. Body poses work well whenever they are "flat" (in front view) but if I slightly distort it to perspective, the results are unpredictable.

I also tried the hand preprocess using a photo, and sometimes it recognizes the hand in the correct pose (again only in front view with slight perspective distortion), but if the fingers are occluded (i.e: behind a driving steering wheel of a car) then I get awful results (multiple hands and extra bodies)

Is there a guideline page as to WHEN to OCCLUDE fingers and HOW to use perspective models correctly? Maybe the finger color code will help as well, I'm generating this in 3D. Please help. Thanks.