“…Although each one uses different representations for shapes and different approaches for part embeddings, they all treat shapes that contain straight lines, circular arcs, or free-form curves or all, equally such that the shape representations become invariant to constituting analytical forms. They either use some registration points as vertices to shape graphs to resolve structure and perform part matchings for graph edges (Keles et al, 2010; Grasl & Economou, 2013); use pixel-based representation and image processing algorithms (Jowers et al, 2010); use weighted representation of shapes and evolutionary optimization algorithms (Keles et al, 2012; Keles, 2015). In all these approaches, shapes are not treated as a composition of a set of pre-defined analytical forms but treated as continuous entities.…”