Motivated by pencil-and-paper analytic drawing, we have described
an approach to constrained inference of 3D lines and curves from
single-view sketches. Our pure-inference interface supports creation
of 3D curve networks comparable to those demonstrated in
recent works such as ILoveSketch [Bae et al. 2008], but without the
need to specify a drawing mode or manipulate the camera for each
curve. By inferring constraints from a well-defined 3D scaffold, it
is more likely that the curve is in the right spatial location, partially
mitigating the effects of drawing skill and perceptual errors.
Although artists readily understand that scaffolds must be assembled
to draw curves, we have found that learning how to construct
a suitable scaffold often requires some geometric ingenuity.
The helix scaffold in Figure 10 provides a good example - once
known, it can be constructed quickly, but it was only discovered
after several failed attempts. The process laid out in design drawing
books [Ching 1997] for blocking out or massing a shape before
adding curves is highly relevant when using our interface. However,
freehand shortcuts are often taken on paper, so even designdrawing
experts will have an adjustment period as they learn how
149:8 • R. Schmidt et al.
ACM Transactions