Hongbo organized a seminar on the Siggraph 2014 papers. I was responsible for presenting the shape analysis section. Here are the slides.
Convert the facade splitting to a shortest grammar problem, and solve it by dynamic programming.
Relation of relations (symmetry & regularity). Use graph matching to find correspondences. Fancy enough.
Instead of studying the 'wind', they study how the 'wind' interacts with the environments. There is a good point in it.
Robustness comes at a cost.