The sketching interface is emerging as an alternative modeling method. Users draw 2D lines on the screen; the system then generates a 3D model automatically, inferring missing depth information. Sketching interfaces for 3D scenes consisting of simple primitives were first introduced in the SKETCH system, 18 allowing users to perform complicated 3D editing operations in a single camera view by combining heuristics. A similar approach is used in commercial systems (such as Google’s SketchUp). However, these systems are designed for sketching simple shapes defined by relatively few parameters. Designing them requires specialized training.

Our group at the University of Tokyo developed the Teddy system11 to address this problem, allowing users to quickly generate interesting 3D freeform models (such as creating a teddy bear by drawing the silhouette of the desired shape) (see Figure 1). The user’s strokes are in red; the system infers and draws everything else. The user first draws the silhouette of the base primitive, and the system generates the corresponding 3D geometry. The user then draws a stroke across the model, and the system cuts the model at the line. The user can also add parts to the base model by drawing two strokes; Figure 2 shows several 3D models created this way.

We do not expect Teddy to replace traditional 3D modeling tools. Rather, it will create new 3D modeling applications that are useful to nonexperts, including children, who want to play with 3D graphics for fun. Introduced at the SIGGRAPH conference in 1999, Teddy is used in several current commercial video games to permit players to create their own characters. Using it is a useful way for experts to express their ideas quickly in early design phases. A commercial 3D modeling package, Shade (available only in Japan, http://shade.e-frontier.co.jp/), includes an extension to Teddy as a plug-in for generating rough sketches. Finally, and most important, Teddy is useful for communicating 3D concepts face to face. In a classroom, for example, a teacher could quickly draw a model of bacteria, showing its cross section to explain its internal structure. In a hospital, a medical doctor could draw a model of a stomach to help explain a patient’s stomach disease.

In 2003, to test the idea, we conducted a trial in a high school geography class in Chiba, Japan. Teaching 3D concepts (such as mountains and valleys), a geography teacher would have difficulty explaining them using traditional 2D media like a blackboard. Sketching in 3D can help address this problem. A convincing example is the teaching of contour lines using the Teddy system (see Figure 3) in which the teacher first shows a 3D model of a mountain, then draws several horizontal lines in the side view, saying the lines indicate equal height intervals. The teacher then changes the viewpoint to show the mountain and the lines from the top. This way, students understand the relationship between the closed lines on the map (contour lines) and the 3D geography, not just mountains, ridges, and valleys.

clothing manipulation
3D characters must also be dressed
properly. The computer-graphics re-
search community has actively in-
vestigated the physical simulation of
cloth, today producing realistic cloth
simulations. However, the initial sim-
ulated-cloth configuration must be
set manually, and the user interface
for manipulating cloth is primitive. A
typical approach is to place rigid cloth
patches around the target body, com-
bining 3D translation and rotation be-
fore starting the simulation—a tedious
process. Moreover, users have difficulty
changing the way the garment is worn
once they’ve placed it on a character.
Standard systems allow users to freely
move individual vertices through direct
manipulation, but it causes a large local
distortion (stretching), making it diffi-
cult to achieve global movement.

figure 4. users draw marks on the character and cloth; the system then places the cloth on the character.

figure 5. Dragging the cloth onto the character: left, before dragging; center, the result of traditional vertex dragging; right, the result of our clothing-manipulation method.