As a non-scientific exercise in qualitative comparison, I’ve assembled here some 3D models made in different photogrammetry programs (VisualSFM, ReCap 360, PhotoScan) from the same source photos. Hopefully this illustrates both the different (and similar) kinds of reconstructions you get with different programs, as well as showing “more faces != better than.” The PhotoScan reconstructions for example are done on “Medium” quality settings in order to speed up processing, and higher face count models could probably be reconstructed there (without much more real increase in “quality” depending on the intended application). Presumably merging together the output of different programs could allow some visualizations of the differences in reconstruction, or yield a better combined model in some instances.
Louvre Pseudo-Seneca
VisualSFM: 1.4M faces, 718k vertices
Louvre Pseudo-Seneca (VisualSFM) by ryanfb on Sketchfab
ReCap 360: 69k faces, 34k vertices
Louvre Pseudo-Seneca (Recap360) by ryanfb on Sketchfab
PhotoScan: 100k faces, 51k vertices
Louvre Pseudo-Seneca (PhotoScan) by ryanfb on Sketchfab
Louvre Caligula
VisualSFM: 967k faces, 483k vertices
Louvre Caligula (VisualSFM) by ryanfb on Sketchfab
ReCap 360: 121k faces, 60k vertices
Louvre Caligula (ReCap360) by ryanfb on Sketchfab
PhotoScan: 195k faces, 98k vertices
Louvre Caligula (PhotoScan) by ryanfb on Sketchfab
Borghese Gladiator
VisualSFM: 1.3M faces, 632k vertices
Borghese Gladiator (VisualSFM) by ryanfb on Sketchfab
ReCap 360: 148k faces, 74k vertices
Borghese Gladiator (Recap360) by ryanfb on Sketchfab
PhotoScan: 60k faces, 30k vertices
Borghese Gladiator (PhotoScan) by ryanfb on Sketchfab
Louvre: L’Enfant à l’oie
VisualSFM: 113k faces, 56k vertices
Louvre: L'Enfant à l'oie (VisualSFM) by ryanfb on Sketchfab
ReCap 360: 46k faces, 23k vertices
Louvre: L'Enfant à l'oie (ReCap360) by ryanfb on Sketchfab
PhotoScan: 60k faces, 30k vertices