December 2014
Beginner to intermediate
154 pages
4h 7m
English
In the last section, we introduced multiple-view distance metrics for pairwise 3-D object comparison. We note that the pairwise object distance is only the low-level comparison that ignores the correlation among different 3-D objects. This distance may accordingly suffer from missed cross-object information. In recent years, learning methods have been superior in many applications, in which semi-supervised learning (SSL) can estimate the structure underlying both the labeled and unlabeled data. SSL is applicable to V3DOR. SSL methods can explore not only the pairwise object relevance but also the joint relevance among multiple objects. In the SSL framework, several recent works have introduced ...