Over the past couple of days, I've been playing around with using a Kinect and the software that comes in the Kinect SDK to do some 3D scanning at home. It has its limitations and the output is usually a little messy around the edges, but so far I'm happy with the results I've been able to achieve with extremely short scanning sessions.
Each of the following only took a minute or two:
Naturally, this has led to me wanting to use these scans to make some highly-detailed characters. I've never cleaned up 3D scans before, so here's how I'm stumbling through that:
Because people tend to move a little and the Kinect isn't super accurate, I had to do a couple of different scans to get everything I needed to start a character. The first scan had a nasty seam running through the middle of the face. I think this has a lot to do with the stitching and construction method that the programs in the Kinect SDK use.
The Kinect isn't very good at picking out extremely small details (the resolution seems to be in the 1-2cm range), so I wasn't able to capture things like eyelid edges and the shapes of the ear. It does capture enough to get the major forms to a pretty good degree of accuracy, though.
After getting those two scans, I used the face from one and the head and shoulders from the other and stitched them together in Zbrush. After that, I just had to refine some things:
I'm using a combination of Kinect output and high-res photos to make sure most of the important details are there.
The plan going forward is to do a few more scans this time with things like clothing and shoes and then start building some accessories so I can make a full character.
So far, I'm really happy with the Kinect as a basic 3D scanner. The output is extremely fast and it doesn't rely on taking a ton of highly detailed photos like photogrammetry solutions do. It's certainly not as accurate as a professional laser scanning or photo-based system, but it's only $100 and it provides a very solid base to sculpt from.