Hi everyone,
I was looking for a way to compare 3D-scan results. Bodys and faces is what i'm interested in. And while my proposed method won't pass academic standards, it seems easy and practical enough for others to reproduce. That's what I had in mind, because I'd like all of you to contribute with your scanners.
My test-object is a lego plate with different piles. I thought there are enough features that can be scanned and compared.
Here's a picture:

(The rubics-cube is placed the for additional geometry.)
I've used my Kinect to scan it and this is what I've achieved:


The nubs have all been smoothed out, but the single lego stones did show up, rounded but clearly visible.
Then I took my photos to recap. It took two days to process (stuck at 95% most of the time). This is what Autodesk came up with:


The nubs come out a lot better. The larger 4-blocks came out very well, hardly any smoothing.
However, one bridge was filled and the other one has a huge dent. The small 1-blocks partially merged.
So both methods have their strengths and weaknesses.
I'd love to see how your scans compare.