Sketching-out Virtual Humans: From 2D Storyboarding to
Immediate 3D Character Animation
School of Engineering and Design
UB8 3PH, UK
Sheng Feng Qin
School of Engineering and Design
UB8 3PH, UK
David K. Wright
School of Engineering and Design
UB8 3PH, UK
Virtual beings are playing a remarkable role in today’s public
entertainment, while ordinary users are still treated as audiences
due to the lack of appropriate expertise, equipment, and computer
skills. In this paper, we present a fast and intuitive storyboarding
interface, which enables users to sketch-out 3D virtual humans,
2D/3D animations, and character intercommunication. We
devised an intuitive “stick figure?fleshing-out?skin mapping”
graphical animation pipeline, which realises the whole process of
key framing, 3D pose reconstruction, virtual human modelling,
motion path/timing control, and the final animation synthesis by
almost pure 2D sketching. A “creative model-based method” is
developed, which emulates a human perception process, to
generate the 3D human bodies of variational sizes, shapes, and fat
distributions. Meanwhile, our current system also supports the
sketch-based crowd animation and the storyboarding of the 3D
multiple character intercommunication. This system has been
formally tested by various users on Tablet PC. After minimal
training, even a beginner can create vivid virtual humans and
animate them within minutes.
Categories and Subject Descriptors
H.5.2 [Information Interfaces and Presentation]: Graphical
User Interfaces (GUI); H.1.2 [User/Machine Systems]: Human
Information Processing; I.3.7 [Computer Graphics]: Animation.
Storyboarding, human modelling and animation, sketching
interface, character intercommunication.
Since the advent of the first computerized human models  in
1970s by aeroplane and car manufacturers, human modelling and
animation have involved constant effort and extensive research
from computer scientists for many decades. Nowadays, its
application has penetrated into a great variety of fields, such as
industry, military, biomedicine, education, etc. In today’s public
entertainment, virtual beings are playing a particularly remarkable
role, when engaged in 3D games, Hollywood films (e.g. “The
Lord of The Rings”, “Star Wars”, “The Incredibles”, etc) and
multimedia (virtual TV presenters). However, creating 3D
characters and their animations is, by far, still remaining the
domain of professionals. Although impressed by the marvelous
visual impact, ordinary people are rarely given the chance to
participate, due to the lack of appropriate expertise, equipment,
and computer skills.
At present, the typical CG animation pipeline entails the
(modelling, rigging, scene layout, animation, rendering, etc), and
postproduction (composing, video editting, etc) . In general,
creating a CG character animation is an extraordinarily
demanding job, which requires a great deal of time, equipment,
expertise, and often the collaborations among various specialists
and production teams. During the above process, although 2D
storyboarding, as a classical animation tool, is relatively simple
and intuitive for the novice to organise scenes , the subsequent
technical tasks are always too overwhelming and unachievable for
To bridge this gap, we developed an intelligent storyboard, which
enables each one who can draw, to “sketch-out” 3D virtual
humans, and their animations, as well as intercommunication.
Through our interface, users can depict the character actions by
drawing simple stick figure key poses, with graphical motion
paths and timing, which can be automatically reconstructed as
continuous 3D motion. Then, users can “flesh-out” any existing
stick figure with body profiles to portray the appearance of their
imaginative character. The system can automatically “perceive”
the body size (skeleton proportion) and shape (body profile and
fat distribution) from the sketched figure, and transfer it into a
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies
are not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
ACE 06, June 14-16, 2006, Hollywood, California, USA.
Copyright 2006 ACM 1-59593-380-8 /06/0006 ...$5.00.
Figure 1. Sketching out a virtual human and animating it in both 2D and 3D world.
characteristic 3D virtual human model. This resulting skin surface
can be then mapped onto each of the posed stick figures, which
can be further interpolated as a variety of 3D and 2D character
animation. Meanwhile, users can build their own 3D character
and motion library, and animate a population of virtual humans
through motion retargeting and a sketch-based actor allocation in
3D space. Moreover, users are also able to sketch-out the
intercommunication among multiple characters in each story
scene. The system can deliver an immediate 3D scenario, in
which virtual actors are acting and intercommunicating with each
The rest of the paper is organised as follows: Section 2 summarise
the previous works in related areas. In Section 3, we brief our
sketch-based human modelling and animation pipeline first,
followed by some details on stick figure keyframing, sketch-based
motion path and timing control, graphical virtual character
generation, and multi-level animation creation with synthesised
panorama and sound/music. In Section 4, we elaborate on how
users can animate a virtual human population, through motion
retargeting and the sketch-based scene layout. In Section 5, we
present the 2D storyboarding for the 3D multiple character
intercommunication. Section 6 contains some implementation
details and user experiences of the current storyboarding
interface. In Section 7, we conclude this paper with some
discussions and ideas for future work.
2. RELATED WORKS
Human modelling and animation is an essential theme of modern
computer graphics, which has drawn remarkable attentions from
the academic and industrial communities for over thirty years.
The research in this area encompasses the development of
numerous techniques: human
, motion control of articulated figures , skin
deformation , facial animation , task level and behaviour
animation , etc. In terms of the virtual human generation,
three major categories can be identified. They are: creative
, reconstructive , and interpolated . In
general, the purpose of all these current works is to generate 3D
human body models effectively and efficiently to meet the
practical needs of the film industry, 3D game, multimedia, virtual
reality, etc. Although capable of creating highly realistic human
body models with variational appearances and motions, these
approaches generally require the extensive expertise, special
equipment (traditional camera, video camera, 3D body scanner,
motion capture system, etc), and proficient computer skills.
Therefore, regular users are still treated as audiences, and have
rarely been given the chance to create and animate their own
virtual characters, due to the lack of appropriate expertise,
equipment, and computer skills.
body shape modelling
At this time, many researchers have recognised the intuitiveness
and importance of sketching, as a design tool, to bring common
users into 3D modelling and animation world with ease and fun.
Since Teddy ’s birth in the late 1990’s, sketch-based 3D
freeform object modelling becomes feasible and an increasingly
hot research issue. Many works  have been developed
to transfer the user’s 2D freehand drawings into 3D freeform
surfaces (i.e. implicit surface, convolution surface, polygonal
mesh) of stuffed toys, simple clothes, car/furniture models, etc, in
various manners. Sketch-based 3D human body modelling,
however, has still remained a difficult undertaking, which has
rarely been addressed in the past. Since the human being is a very
complex object and our eyes are especially sensitive to the human
figure, current modelling techniques constrained by a spherical
topology as in , are essentially not adequate for
obtaining the plausible results.
In recent years, sketch-based 2D/3D animation is growing rapidly
as an interesting and promising research area. Several sketching
interfaces have been developed for articulated figure animation
, motion doodling , and cartoon storyboarding .
Meanwhile, another corpus of research work have been
specialised in infusing the expressiveness of traditional 2D
animation into 3D animation, through cartoon capture and
retargeting , motion stylisation , and view-dependent
animation . Although impressive, many of these systems
 are only dealing with single and simple character
in each animation, rather than the sophisticated human characters
with realistic appearances and variational actions. While essential
for storytelling, the intercommunication among multiple actors
has rarely been supported. Meanwhile, none of these systems has
delivered a complete picture of the “sketch-based modelling and
animation”, including key framing, figure pose recognition, 2D-
3D surface modelling, and the resulting animations performed by
various levels of characters (e.g. stick figures, 3D mesh models,
2D NPR models, etc.)
3. SKETCHING-OUT SINGLE
3.1 Sketch-based human modelling and
As in , a sketch is essentially a noisy projection of a 3D object
onto an arbitrary plane, and the reconstruction is an inverse
projection of the sketched geometry from 2D back into
3D. In terms of the perception of raw figure drawings, the human
brain can envision their 3D counterparts easily, and even
spontaneously. This is, however, rather difficult to be performed
by computer, especially when provided with a drawing that has
many ‘noises’, such as, foreshortening, contour over-tracing, body
part overlapping, shading/shadow, etc.
To decompose the complexity of the direct 3D modelling and
animation from fully rendered sketches, we have designed a
“stick figure ->fleshing-out->skin mapping” pipeline (as
illustrated in Figure 1). This was inspired by the drawing
sequence recommended by many sketch books and tutorials
. Meanwhile, it principally echoes the prevalent CG
animation pipeline, whilst employed instead in a sketch-based
intuitive way. On this sketching interface, users draw stick figure
key frames first to define a specific character motion. Unlike a
common character choreography, these illustrative sketches can
be automatically reconstructed by the system as immediate 3D
motion (see Figure 1(b)), which remits the additional 3D key
framing job by users. Then, users can choose a single stick figure
for “fleshing-out” (see Figure 1(a)), which is akin to the character
visual appearance design. The system can “perceive” the body
features of the sketched figure, and “pop-it-up” into its
counterpart 3D character directly, which effectively relieves users
of the tough and time-consuming 3D modelling task. After that,
users do not need to perform the character rigging. The system
can automatically “wrap” a single skin surface onto a series of
posed stick figures and interpolate them as the final 3D full figure
animation (Figure 1(c)). Moreover, 2D contour animation (Figure
1(d)) and 2D NPR (Non-Photorealistic Rendering) animation
(Figure 1(e)) could also be delivered with a personalised sketchy
Regarding our current design, the functionalities at different
levels were achieved for different users. Thus they can choose to
make simple stick figures, create delicate 3D surface models, or
explore further to animate these sketch-generated creatures.
Moreover, models could be exported to commercial packages (3D
Studio Max, Maya, etc.) at any level, to be refined by their
powerful function kits.
3.2 Sketching-out 3D stick figure key frames
As shown in Figure 1(a), users can convey a specific character
motion by simple stick figure drawings. In our storyboarding
system, an on-line drawing assistance is provided to help users
maintain the proper figure proportion and foreshortening. For 2D-
3D pose recovery, we developed a “multi-layered back-front
ambiguity clarifier” , which utilises figure perspective
rendering, human joint ROM (Range of Motion), and key frame
coherence to identify the user intended 3D poses. In addition, a
“figure pose checking/auto-correction” routine is offered to
ensure the physically valid poses. Moreover, we support a sketch-
based interactive motion path and timing control (see Figure
8(top)) . Sound/music and panorama (see Figure 5(Top),
Figure 7-9) have been provided to enhance the 3D virtual world.
Alternatively, users may import their own selections. The
resulting 3D stick figure animation is synthesized in VRML,
which can be triggered by a single user click. The technical
details have been presented in .
3.3 Sketching-out variational 3D virtual
3.3.1 Creative model-based 3D body generation
As early introduced, users can depict the visual appearance of
their virtual character through “fleshing-out” a single stick figure
with body profiles. We developed a “creative model-based
method”, which emulates a human perception process to generate
the variational 3D human bodies from 2D freehand sketches,
through morphing a pre-stored 3D generic model. Our generic
model is created from an anatomical image dataset , which
has been encapsulated with three distinct layers: skeleton, fat
tissue, and skin. Figure 2 illustrates this computer-simulated 2D to
3D perceptual process. Since only one generic model is
employed, our current system could not generate a full range of
population (including female, children, elders, etc). This,
however, can be solved in the future by creating a wide range of
morphable template bodies for modelling use. The sketch-based
modelling of human heads/hands/feet is not feasible at this stage,
which is also a common challenge for other related approaches.
Figure 2. Transfer a 2D freehand sketch into the 3D body
model through an automatic system perception and morphing
process: Users draw a 2D figure (S0). The system
automatically retrieves its 3D pose and body proportion, and
performs a rigid morphing on the 3D generic model. The
resulting 3D model M1 is projected into 2D (blue lines in S1)
and compared with the original sketch (black lines) to
evaluate its body fat distribution. M1 is then deformed
through fatness morphing into M2, which is projected (green
lines in S2) and compared again with the 2D sketch to get the
fitting measurements. The final 3D model (M3) is delivered to
users, after an automatic surface fitting and beautification
process (on M2). Users can incrementally refine their 2D
sketches; a similar perception/morphing process is performed
to produce the updated 3D model.
(a) (b) (c) (d)
Figure 3. (a) The input 2D freehand sketch and the 3D model after rigid morphing. (b) Graphical comparison to get the fat
distribution measurements and the fatness morphed model (c) Graphical comparison to get the surface fitting measurements; the
model with and without system auto-beautification (d) Overtracing body contour (right lower torso, and lower legs) to modify an
existing 3D surface model.
3.3.2 Transfer 2D raw sketch into 3D plausible body
As shown in Figure 3(a), freehand figure sketching is essentially
rough and imprecise, which contains various “noises” –
wiggly/overlapping strokes, missing figure contour, asymmetrical
body parts, etc. To turn this sketchy 2D figure into 3D
sophisticated mesh model, sketch clean-up needs to be performed.
Figure 3 illustrates how a freehand figure sketch undergoes
automatic processing and graphical comparison, to be morphed
gradually (both “biologically” and geometrically) into a plausible
3D human body model (details in another paper). During this
process, an “auto-beautification” option is also offered to
regularise an asymmetrical sketch-generated model (see Figure
3(c)). Meanwhile, users can interactively refine the resulting 3D
model by over-sketching its 2D figure profiles (see Figure 3(d)).
Modifications can be made at any time, and on any key frame
sketch, to achieve the updated 3D model. In addition, a post-
processing function is also provided for varying an existing figure
model, by changing its body proportion.
Beyond a flat drawing medium, it will be ideal to provide users
with an interactive and mixed modelling environment, in which
they can sketch 2D figure, “pop it up” into a 3D character, and
incrementally refine it through suggestive contours ,
shading/shadow, etc, in both 2D and 3D. Having realized the
sketch-based figure fast prototyping, the implementation of other
features will be our next challenge.
3.3.3 Generate 2D and 3D virtual human animation
Following the “stick figure ? fleshing-out ? skin mapping”
pipeline, a 3D virtual human animation is accomplished by
wrapping the sketch-generated skin surface onto a series of posed
stick figures, which can be further interpolated via VRML ,
with the associated graphical motion definition. Figure 5(Top)
shows the snapshots of the 3D animation specified by Figure 4,
which provides an insight of our virtual human sketching
interface, and its intuitive graphical tool kits.
While inspiring to “pop-up” 3D virtual beings by freehand
“doodling”, it is also an amusing experience to animate the 2D
sketchy figures, like 2D cartoon storyboarding. This is different
from the traditional cel animation, and users do not need to render
each key frame, once a single key figure fleshed-out.
Figure 4. The user is sketching-out a virtual human and its
motion on a Tablet PC.
Figure 5. (Top) A sketch-generated 3D dancing character.
(Bottom) The 2D NPR animation played on the sketching
interface in a doodled countryside view.
Figure 5(Bottom) shows the snapshots of a 2D NPR animation
(derived also from the storyboarding on Figure 4) in a doodled
4. CREATE AND ANIMATE VIRTUAL
The techniques for animating computer-generated three-
dimensional crowds were greatly refined during the late 1990,
starting with feature films like “The Lion King” and “The
Hunchback of Notre Dame”, and in conjunction with films like
“Prince of Egypt”, “A Bug’s life”, to today’s “The Lord of the
Rings” and “Star Wars” . Regarding the exceptional effect of
the crowd animation, in our storyboarding system, we offer a
simple sketch-based function for users to animate a population of
sketch-generated 3D characters.
In our system, users can either create the assorted virtual humans
and their motions by 2D sketching, and store them into the
interface-embedded character and motion library; or choose the
ones they like from the system provided actor and action lists.
Since the 3D motions of some individuals might be similar in a
group animation, we offered a motion retargeting function, which
can adapt a single action onto multiple characters with various
appearances. To specify the character locations on the 3D floor,
users can draw simple crosses or circles (see Figure 7(a)) onto its
top view plan, to indicate the starting positions (body root XY
value) of the crowd individuals. Here, we presume that each
character’s feet are touching the ground (default as flat) at the
beginning of the motion. Users can define the shape of the floor
either by selecting from the provided patterns, or by customising
some existing ones.
In brief, the procedures for creating a group animation (in our
Create/select a 3D virtual actor.
Create/select a 3D motion.
Motion targeting to animate the selected character.
Draw 2D landmark (cross/circle) to locate the character
onto 3D virtual floor.
Repeating step 1-4 for each group individual, until the
whole animation set-up finished; play to view the group
animation result in VRML.
Figure 6 below shows the sketches and a range of variational 3D
human bodies, which were created by various users in our user
test (details in Section 6). Figure 7(a) illustrates the sketch-based
character allocation. Figure 7(b) shows the group animation
result, in which a population of virtual humans from users and
some previously illustrated figures are playing Chinese Kungfu
together in 3D virtual world with music (see also the attached
video clips). Here, a single Kungfu motion was applied onto each
of the group individuals. In Figure 9, a crowd animation with
intercommunication and variational actions is presented.
5. 2D STORYBOARDING FOR THE 3D
actors/actresses is usually vital to transfer the drama and emotion
of the story and characters. Meanwhile, some psychological
researchers have concluded that more than 65 percent of the
information exchanged during a face-to-face interaction is
expressed through nonverbal means . In our storyboarding
interface, users can draw multiple characters in each keyframe to
show the flow of interactions, like defining the visual camera
shots (see Figure 8(Top)). To draft the character motions, users
can sketch either fully rendered figures, or only simple stick
figures to illustrate prompt ideas. Meanwhile, users can also
annotate their drawings with script dialogue, or some other
notes/symbols to assist the storytelling. The system can
reconstruct the 3D motion for each character sequentially from
the 2D storyboards. For the motion path editing, users can either
sketch-out the motion curves separately for each individual, or
specify it for just a single actor. The system can automatically
estimate the associated motion information for the other
characters through analysing their relative interface locations
(depth relationships not included). Similarly, users can sketch-out
the timing for each character, or define a single one and apply it
onto the other story actors. After the 2D storyboarding and the 3D
reconstruction, the final 3D animation with interactive characters
can be displayed via VRML after a single user click. Figure
8(top) shows the 2D storyboards of two characters Kungfu
fighting with each other, with some associated annotations and
motion curves. Figure 8(bottom) presents the snapshots of the
synthesised 3D fighting motion. Meanwhile, users can also create
the group animation of interactive characters through the same
means as described in Section 4. The default local origin for each
character set is the body root of the leftmost character on the
corresponding first keyframe sketch (see Figure 8(top)). Figure 9
shows an interesting animation of a crowd of virtual humans and
even stick figures fighting with each other in 3D virtual world.
Among them, some characteristic figure actions (appeared in
Figure 1 and 5) could be spotted to show the individualities
among the collective behaviours.
2D storyboarding, the intercommunication among
6. IMPLEMENTATION AND USER
Our prototype modelling and animation system is implemented by
Microsoft Visual C++, MATLAB, and VRML. This system has
been tested on a variety of input devices: electric whiteboard,
Tablet PC, as well as a standard mouse.
On the completion of our storyboarding system, we conducted a
formal user evaluation to assess its usability and functionalities
with various users (including
undergraduate/postgraduate students, and a 12-year-old boy),
through performance tests, sketching observation, and user
interviews. After a short tutorial, users rapidly learned the
storyboarding process, and began sketching-out their own virtual
humans and animations (on Tablet PC) within minutes. In
general, the overall average time for creating a complete stick
figure animation and a complete full figure animation (both
containing 3 keyframes) is 6.27 minutes and 6.75 minutes
respectively. It is in remarkable contrast with the commercial
packages, which requires usually dozens of minutes and several
hours respectively for an animator, to create the similar
articulated and full character animation from scratch.
S1 S2 S3 S4 S5 S6 S7 S8 S9
Figure 6. A variety of 3D virtual humans and the original drawings by different users: artist (S3), design student (S4), animator
(S6, S7), graduate students (S1, S2, S5, S9), and a child (S8). The inner contours on some sketches (S3, S4, S6, S7, S9) are added
after the 3D model generation, which shows user’s intention to depict the detailed surface shape through more rendering forms.
Figure 7. (a) 2D top view plan with the denoted character locations (S1-S9 and F1-F2 are the indices of the characters, which
appeared in Figure 6 and Figure 1,5), and the corresponding 3D result. (b) Kungfu group animation with music and background
(the pink ball on the floor is a touch sensor to trigger the music).
Figure 8. (Top) The Kungfu fighting storyboards with the associated annotations and motion curves (Xk and Yk are the X and Y
trajectory curves for the kicker. Yd is the Y curve for the defender, who’s path x value can be automatically estimated. The motion
timing was act-out by the curve drawing speed.) (Bottom) Two sketch-generated 3D characters are fighting with each other.
Moreover, this modelling/animating by sketching approach is
proved by users to be easy to learn and use, and entertaining to
play with. Figure 6 shows the sketches and the variational human
bodies created by users during the test, which have been
integrated into two group animations, shown in Figure 7 and 9.
7. CONCLUSIONS AND FUTURE WORK
In this paper, we have presented a fast and novel storyboarding
interface for sketching-out 3D virtual humans, 2D/3D animations,
and character intercommunication. Human modelling and
animation is a recognisable challenge and a labour-intensive task,
which has been, until now, confined to the domain of
professionals. This research draws on the people’s existing
drawing skills and the intuitiveness of 2D storyboarding as a
design tool, to enable ordinary users to create and animate their
own living characters, with ease and fun. Our main contributions
in this work are: 1) we devised an intuitive “stick
pipeline, which realises the whole process of key framing, 3D
pose reconstruction, virtual human modelling, motion path/timing
control, and the final animation synthesis by almost pure 2D
sketching; 2) we investigated an easy and fast approach, which
enables the sketch-based crowd animation and the storyboarding
for the 3D multiple character intercommunication. On the success
of the current system, we envisage its applications in human
modelling and animation, CG animation/film making, cartoon
storyboarding, interactive game (on Internet, home PC, or mobile
devices), virtual reality, education, etc.
The animation system presented here is not meant to be a
substitute for the array of professional animation tools and
techniques that are commonly used in film and game production.
Instead, our storyboarding system provides an alternate and more
accessible means for ordinary users, to create the simple and
entertaining animations in various forms. Meanwhile, this system
can be also utilised as a fast prototyping machine, from which the
sketch-generated 3D character models and animations can be
created and imported into commercial tools, to be refined by their
powerful function kits.
In the future, we are going to extend our system to allow the
generation of a full range of population, including females,
children, elders, etc. We will realise an interactive 2D and 3D
drawing environment, and adopt more rendering forms (i.e.
suggestive contours, shading/shadow) to depict subtle surface
features. We will investigate the appropriate means to detect and
avoid the collisions in a crowd animation. Meanwhile, it is also
our future work to further enhance the system functionalities,
through enabling the storyboarding of character expressions, and
the writing-based phonic dialogue creation.
 Isaac, V.K. The art of 3D computer animation and effects. 3rd
edition, John Wiley & Sons, Inc., Hoboken, New Jersey,
 Fox, B. 3ds max 6 Animation: CG Filmmaking from Concept
to Completion. McGraw-Hill Osborne, California,
 Hoshino, J., and Hoshino, Y.: Intelligent storyboard for
prototyping animation. In Proc. IEEE Int. Conf. On
Multimedia and Expo (2001), Conference CD-ROM FAI.03.
 Wilhelms, J., Van Gelder, A. Anatomically based modeling.
In Proc. SIGGRAPH ’97 (1997), 173-180.
 Magnenat-Thalmann, N., and Thalmann, D. The direction of
synthetic actors in the film Rendez-vous à Montréal. IEEE
Computer Graphics and Applications 7,12 (1987).
 Hilton, A., Beresford, D., Gentils, T., Smith, R., and Sun, W.
Virtual people: capturing human models to populate virtual
worlds. In Proc. Computer Animation, IEEE Press (1999),
 Allen, B., Curless, B., and Popovis Z. The space of all body
shapes: reconstruction and parameterization from range scans.
In Proc. SIGGRAPH 2003 (2003), 587-594.
 Anguelov, D., Srinivasan, P., Koller, D., Thrun, S., and
Rodgers, J. SCAPE: Shape Completion and Animation of
People. Proc. SIGGRAPH 05 (2005), 408-416.
 Grochow, K., Martin, S. L., Hertzmann, A., and Popovic, Z.
Style-based inverse kinematics. ACM Transactions on
Graphics (TOG) 23, 3 (2004), 522-531.
 Weber, J. Run-time skin deformation. Intel Architecture
Labs, http://www.intel.com/ial/3dsoftware/index.htm (2006).
 Ekman, P., and Friesen, W. Facial action coding system.
Consulting Psychologists Press, Palo Alto, 1978.
 Bruderlin, A., and Calvert, T.W. Goal directed, dynamic
animation of human walking, Proc. SIGGRAPH ’89,
Computer Graphics 23, 3 (1989), 233-242.
 Turner, R., and Gobbetti, E. Interactive construction and
animation of layered elastically deformable characters.
Computer Graphics Forum 17, 2 (1998), 135-152.
 Fua, P. Human modeling from video sequence. Geometric
Info Magazine 13, 7 (1999), 63-65.
 Igarashi, T., Matsuoka, S., and Tanaka, H. Teddy: a
sketching interface for 3D freeform design. In Proc.
SIGGRAPH ‘99 (1999), 409-416.
 Alexe, A., Gaildrat, V., and Barthe, L. Modelling from
sketches using spherical implicit functions, Proc. 3rd
international conference on Computer graphics, virtual
reality, visualisation and interaction in Africa (2004) 25-34.
 Karpenko, O., Hughes, J. F., and Raskar, R. Free-form
sketching with variational implicit surfaces. Proc.
Eurographics 2002 21, 3 (2002).
 Wang, C.C.L., Yuen, M.M.F. Freeform extrusion by
sketched input. Computers and Graphics (2003), 255-263.
 Davis, J., Agrawala, M., Chuang, E., Popovi?, Z., and
Salesin, D. A sketching interface for articulated figure
animation. In Proc. Eurographics / SIGGRAPH Symposium
on Computer Animation (2003), 320-328.
 Mao, C., Qin, S.F., and Wright, D.K. A sketch-based gesture
interface for rough 3D stick figure animation, Proc. of
Eurographics Workshop on Sketch Based Interfaces and
Modeling (2005), 175-183.
 Thorne, M., Burke, D., and van de, Panne, M. Motion Download full-text
doodles: An interface for sketching character animation.
ACM Transactions on Graphics 23, 3 (2004), 424- 431.
 Bregler, C., Loeb, L., Chuang, E., and Deshpande, H.
Turing to the masters: motion capturing cartoons. ACM
Transaction on Graphics 21,3 (2002), 399-407.
 Li, Y., Gleicher, M., Xu, Y-Q, and Shum, H. Stylizing
motion with drawings. Proceedings of 2003 Symposium on
Computer Animation, 2003.
 Chaudhuri, P., Kalra, P., and Banerjee, S. A system for
view-dependent animation. Eurographics 2004 23, 3
 Lipson, H., and Shpitalni, M. Correlation-based
reconstruction of a 3D object from a single freehand sketch,
In Proc. AAAI Spring Symposium Series - Sketch
 Tiner R. Figure drawing without a model. David & Charles
plc, Glasgow, 1992.
 Li, W. Figure drawing: Basic pose and construction.
 National Library of Medicine. The Virtual Human Project,
 DeCarlo, D., Finkelstein, A., Rusinkiewics, S., and Santella,
A. Suggestive contours for conveying shape. ACM
Transaction on Graphics 22, 3 (2003), 848-855.
 Ames, A. L., Nadeau, D. R., and Moreland, J. L. VRML 2.0
Sourcebook. 2nd edition, John Wiley & Sons Inc., 1997.
 Argyle, M., Bodily communication. Methuen & Co., New
Figure 9. A crowd of variational virtual humans and stick figures are fighting with each other in 3D virtual world. Two of them (the 5th
from the left, and the right most figures) are acting differently, which shows the individualities among the collective behaviors.