The other three datasets (OFFICE 1, OFFICE 2 and OFFICE 3)
represent different office environments which are composed of
corridors with attached rooms. The highly anisotropic shape of the
corridor in OFFICE 1 (Fig. 13(c)) is reconstructed correctly and
separated from the neighboring room. Thanks to our post-processing
step, the over-segmentation problems that affected our original
pipeline [3] are solved. OFFICE 3 (Fig. 13(e)) represents a larger
environment composed of several rooms, all attached to a central
corridor. While OFFICE 3 exhibits a clear arrangement of the
rooms, the environment in OFFICE 2 (Fig. 13(d)) has a more
complex and irregular structure, lacking a single central corridor
and containing some empty space completely surrounded by other
rooms. Despite these challenges, all the rooms are correctly
detected and reconstructed.