I did not mask it. No need to when using black (self adhesive) velvet as a background. But perhaps I will be corrected on this when I get more experience in 3DGS. The 885 VPs are a breeze with this robot. Captured in 2h35m. I was very surprised to see that RS was capable of aligning all cameras 100% within 20 minutes, despite using focus stacking (and thus in fact having multiple VPs per image). I have build this robot to be able to do a calibration scan of a calibration object that is shot from the mid-frame of every stack at every VP with the exact same VPs (with sub-pixel accuracy) and then transfer these camera positions to the object scan (I figured that this could be helpful for low-feature object - to avoid having to use doodle objects). That's why i designed and build this robot to be very accurate. It is a 4 axes setup with fully interactive 3D UI and live-view directly from the camera. All editing of the scan shape is done visually and in 3D space interactively . The robot has one elevation (EL) axes (Harmonic drive ~ 2 arcsec accuracy) that carries a proximity (PROX) axes (ball screw slide rails) where the camera rides on (focus stacking axes) and one height axes (ball screw rails) carrying the stage with the azimuth axes (AZ - turntable or pin) to turn the object in a non-dynamic gravity field, thus making this robot also suitable for soft-body specimen. I am still aiming at a reduction of the transparency of a 3DGS. It thought that increasing the number of VPs would result in better opacity, but I can't say I can see much of a difference. There is still a lot to learn and experiment with.. Thanks for your beautiful and inspirational work!