Traditional 3D modeling tools (e.g. [Autodesk 2009]) require users
to learn an interface wholly different from drawing or sculpting in
the real world. 2D drawing remains much easier than 3D modeling,
for professionals and amateurs alike. Professionals continue
to create 2D drawings before 3D modeling and desire to use them
to facilitate the modeling process ([Thorm¨ahlen and Seidel 2008;
Tsang et al. 2004; Eggli et al. 1997; Kallio 2005; Dorsey et al. 2007;
Bae et al. 2008]). Sketch-based modeling systems, such as Teddy
[Igarashi et al. 1999] and its descendants, approach the 3D modeling
problem by asking users to sketch from many views, leveraging
users’ 2D drawing skills. In these systems, choosing 3D viewpoints
remains an essential part of the workflow: most shapes can only be
created by sketching from a large number of different views. The
workflow of these systems can be summarized as “sketch-rotatesketch.”
Because of the view changes, users cannot match their
input strokes to a guide image. Moreover, finding a good view for
a stroke is often difficult and time-consuming: In [Schmidt et al.
2008], a 3D manipulation experiment involving users with a range
of 3D modeling experience found that novice users were unable to
complete their task and became frustrated. These novice users “po