Viewpoint Selection for Photographing Architectures

Enabling Viewpoint Learning through Dynamic Label Generation

“…Accordingly, researchers have turned to automation to extract user preferred views from existing data. This can be done by computing similarities between images and candidate viewpoints of 3D models [Liu et al 2012], through a clustering strategy to recover viewpoints for architectures [He et al 2017], or by using a CNN viewpoint estimator to retrieve viewpoint probabilities from publications for medical data [Shi and Tao 2019]. In contrast to these approaches, we propose to combine a data-driven elevation distribution estimation from images with information theoretical viewpoint quality measures.…”

“…Many quality metrics have been developed to aid in the automatic selection of optimal viewpoints on 3D models. The applications range from obtaining vantage points for capturing stills in architecture [He et al 2017], to initial camera positioning for complex scene inspection [Heinrich et al 2016;Meuschke et al 2017;Secord et al 2011;Song et al 2014], camera control [Lino and Christie 2015] and recommendations for scientific visualization [Yang et al 2019].…”