They use stereopsis for coarse scale depth and photometric stereo (three directions from the looks of it) for finer scale structures. And they seem to be using some tracking target to compensate for motion between these captures. Not a bad idea per se, but I don't think their numbers are particularly remarkable.
I'm not aware of any 3D capturing technique that captures an object "from all sides", unless it's comprised of multiple individual scanners who's data you then stitch into a single model, or a moving scanner (relative to the object's reference frame, so the object could be the one moving), in which case you're really building the model out of lots of tiny scans at different positions (e.g. sheets). In principle, either of these are more or less orthogonal to the choice of scanner. You could do it with this scanner; you'd just put them all in a box and calibrate the extrinsic parameters using some reference object.
The only things that I can think of that could be remotely considered scanning from "all sides" would be something that penetrates the object, like an x-ray CT scanner, ultrasonography or something of that sort, but that would be stretching it.