New user struggling to get cameras to align

[totalitarian]

Completely new user to both photography and scanning.

This is my collection of photos

https://www.dropbox.com/sh/5l6drttilg6dp6l/AAAO1HjahB05ku8aMpskuxe8a?dl=0

I tried following the workflow but got stuck on the align photos section. I can only assume is is due to the following

1) The images/lighting is not adequate
2) The turntable is confusing the software.

I read in the manual about masking. I'm a little confused what to use for the background image though. Do I need the background image to include the turntable or not? When I tried using masking from background option, the results were a little poor. Several areas were not masked correctly, particularly around the subject's arms. I guess it doesn't help the hands are nearly the same colour as the turntable.

The other confusion i'm having is that I tool a set of photos at roughly 0 degrees to the subject, then another at a slightly higher height. Do I throw them all into the same chunk?

Hope you can help as i'm keen to master this subject.

[dmacd]

If all you want to model is the figure, then you should definitely try using masks! If you mask out the background (use the magic wand, and then edit it if need be), then the program should be able to align the photos much more accurately.
Also, those photos can all go in the same chunk! When Photoscan aligns the photos it will recognize where to put them all.

[totalitarian]

Thanks, manually masking did the trick. Is there any technique to avoid having to mask? Took me just over an hour

Not quite sure what I need to improve but even after masking, the model comes out a little unrefined 

[SAV]

Hi totalitarian,

Try this brand-new volumetric masking feature (available since version 1.4) in order to minimise the masking workload.

http://www.agisoft.com/index.php?id=48

Regards,
SAV

[totalitarian]

Thanks, I did see that that but the workflow confuses me as it mentions to do the masking during the mesh stage. I need to mast at the align photos stage otherwise my cameras don't align properly... or have I got confused?

[Alexey Pasumansky]

Hello totalitarian,

The masks should be applied to all the scene elements that do not move accordingly to the object.

To speed up the masking procedure you can use an image of the background only from the same camera position, but without the object. So you can try to use Import Masks option -> From Background and use static filename in the template field, like "background.JPG" in your case.

Also masking using magic wand works quite effectively on your images.

[Alexey Pasumansky]

Hello totalitarian,

You can try to follow new masking functionality with your dataset:
http://www.agisoft.com/index.php?id=49

We have masked three images from your set: background.JPG (completely), background2.JPG (completely) and _DSC3926.JPG (accurately, as shown in the attachment). Then aligned the photos using Tie point masking option, generated mesh using new method (from depth maps) and then refined the mesh using 10 iterations and smoothed it once. The screenshots of the resulting model are also attached.

[totalitarian]

Thanks! I'll give that a go