Hi, sorry I'm late to this thread, but I wanted to share my solution. I made a pegboard by drilling a grid of holes into a piece of wood. Then I use very long nails through the bottom to hold up the subject. Sorry the data is on a different service, but here's the model.
This works in Photoscan, too. I've actually never bothered masking photographs in Photoscan or in any other photo-modeling software, and have always gotten usable results. If the photos are good, you'll nearly always get a workable result.
Reviving an old thread here but I like that pegboard setup. I use a spike for these vegetables I did but obviously that's not a very good idea for say, a skull! XD Otherwise I just use the turntable method with two chunks (top/bottom) and a masked out turntable because I don't want to damage anything, although I've not done much of this at all. I had it work excellently for a cube like object (http://d2f0ora2gkri0g.cloudfront.net/bkpam2167114_photoscanexample_2.jpg) but fail for a skull I scanned, makes me wish we did have levitation technology...
OK, here's one idea that came to me while photographing an object on a light box.... scanning on a lightbox. It solves two key problems... back lighting to remove shadows and providing a featureless background (when slightly overexposed). Our new A2 lightboxes are a bit of a pain for most applications, but in this case the extra size made it possible to get down low enough and still have a pure white background. An old Xray lighbox would do the job. This could work on a turntable but in this case I just manually turned the object and then flipped it upside down for the bottom. No masking (with Reality Capture anyway)
I have just been rotating the object itself by hand and then masking out my fingers. For smaller objects I use a solid background of a contrasting colour to the object. This makes auto masking very easy.
I used a hand rotated approach with these pearlware sherds. They are also glossy so I had to diffuse the light.
One chunk, hand rotated, ~1cm thick in the thick body. The flake mended on the end is at most 2-3mm thick.
masks, Masks and MASKS! Especially if we talking about Agisoft Photoscan because without them it produce too much weird noise. Photoscan have good tools for automatic creation mask. You took 1 photo without object and it automatically create mask use this clean photo depend on difference between photo with object and without.
And take photos how gerpro whrote:
If you have camera with GPS, clean GPS tags, especially if this is ContextCapture, GPS positions have priority in calculation camera placement.
So recently for this scan of a porcelain skull I photographed (with lights and a green screen) the more complex bottom with the middle so they would merge much better being in the same chunk and batch of photos. I should say though plasticine really helped out here for holding the thing too, I could just delete out all of the dense cloud with the plasticine showing and keep what I needed, then merge both halves
I know that I am very late to this conversation, but I have recently learned a method that a student at UNC Chapel Hill has worked with. I use a scanner stand (similar to your wood prop @abbyec) on a turntable. For a lithic point, I used a white backdrop and a black, rubber nodule and platform to hold the artifact vertically. Without changing the setup of the camera I rotated the projectile point 90 degrees after every set of photos. In Agisoft, for all but one artifact position of photos, I used the [Smart Scissors] to follow around the edges of the artifact and masked everything outside. I kept them all in one chunk and aligned the photos. If successful, I masked the remaining position photos and built the dense cloud.