This is a prototype system for taking mobile-phone video - typically fairly low resolution - and using it to reconstruct geometry. At the moment the process takes about an hour to calculate and send back the mesh. We are currently building the pipeline to run on GPU cluster instances in the cloud, which should speed things up.
Credit should also go to Yasutaka Furukawa for the PMVS; the dense reconstruction part of the pipeline.
I'll set up the email for all once I can figure out a way to get cheap cloud processing.
I'm keen to hear your thoughts!