|
|
||
|
Motivation and Goal Few people learn how to
sketch at school or during their professional education, but still, most
everybody is able to draw a sketch or to understand a sketched scene that was
drawn by somebody else. This is sometimes even more astonishing, because
people’s sketching habits and techniques can differ considerably from one
person to another. Nevertheless, we believe that sketches contain reoccurring
structures, pattern, and symbols and that there are certain drawing
strategies that are commonly used to generate sketches. Hence, the primary reason
for studying the sketching behavior of people is that a thorough
understanding of the way people sketch is an essential requirement for the
development of methods and techniques that allow an automated interpretation
of freehand sketches. This survey was, therefore, designed to provide the
required base knowledge for Spatial-Query-by-Sketch and similar sketch-based
applications. |
|
|
|
Methodology In the scope of this
survey we have analyzed a total of 91 sketches from 31 people with various
cultural, educational, and professional backgrounds. The age and gender was
quite heterogeneous as well, with an age ranging from 25 to 57 years,
including 11 female and 20 male subjects. The task for each subject was to
draw three sketches based on written instructions and to answer some
sketching related questions. We have analyzed every
sketch manually on an object by object base (total 1208 sketch objects) and
fed the results into a database. Beside sketched objects we have analyzed
also their neighborhood relations and eventual object annotations. The
subsequent interpretation of the survey is, therefore, based on general
observations, an analysis of the written questions asked in the survey and
most of all a huge number of database queries of our sketch database. For a comprehensive description of the methodology and the
results of the survey, |
||
|
Summary of our Findings There is no space to take
every detail of our findings concerning peoples sketching habits and pattern
into this summary. Instead we have compiled a reasonable set of observations
that we think are essential in the first place when dealing with geo-spatial
freehand sketches. For a more detailed discussion we have to refer to the
technical report itself. During our examinations,
we have observed that people keep the number of objects that they use to
describe a spatial situation relatively small. Depending on the complexity of
a sketch we experienced in average between 12 and 17 objects per sketch. We
have also experienced that most subjects seem to favor artificial object,
such as roads, buildings, or other human built monuments over natural
objects, such as vegetation or topography in their sketched representations.
An exception are some large scale natural objects with clear boundaries, such
as rivers, lakes, or the sea. Objects of type forest, hill, or valley on the
other hand are quasi non-existent. In this context we have observed the
general tendency that people tend to represent geo-spatial situations in a
pure two-dimensional, map-like manner without taking any topographic features
into account. If the third dimension is considered then this is limited
mostly to an object or to a group of objects, such as the front view of a
house or the silhouette of a town. People keep the shape and
structure of objects simple, in that they use only a few strokes to draw an
object and in that they tend to use symbols to represent real world objects.
Such, it is no surprise that boxes and simple lines are the preferred forms
for representing object in a sketch. We observed also that objects are in
general relatively carefully drawn, so that for instance boxes are closed and
lines representing streets tend to meet but not to cross where they are
supposed to. Because of their simplicity, sketched objects taken out of
context do frequently have no own meaning. We found two ways how people
attached a meaning to these objects. The first uses the context of an object
and the ability of people to infer a meaning, while focusing on a group of
objects within a particular spatial configuration. The second possibility is
to annotate sketched objects verbally or by writing. We could not record
verbal annotations but we found that almost two third of all sketched objects
of our survey had some kind of attached annotation. Annotations in our survey
were in general simple words or ellipses, while phrases were rarely used. We were surprised about
the use of metric information in sketches as we have observed that people
seem to rely primarily on topology and the composition of objects to describe
a geo-spatial scene. Explicit distance indications were almost non-existent
in our survey. However, some implicit forms of distance indications, such as
the relative closeness of objects seem to be important. Direction and orientation
of objects and object groups seem to play an essential role in sketches. We
found that a great number of sketched objects had an assigned or at least
deduced direction associated and that many objects were linked to other
objects through parallel or rectangular relationships. The preferred
orientation for objects and the sketch itself is along the North-South or
West-East axis of the sketching device. But, despite our subjects’ general
preference to align their objects according to these two cardinal
orientations, we noticed a slight upwards deflection to the upper right for
many objects. The context and the
immediate spatial neighborhood of an object in a sketch are crucial. We have
found that sketched objects are typically connected to each other via
neighborhood relations of type disjoint, overlap, or meet, while other
topological relations indicating containment are only sparsely used.
Considering their location we have observed that objects drawn in temporal
sequence are typically closer to each other than objects that are out of
sequence. However, we have found also other sketching strategies, where
people tend to sketch a sequence of objects based on their thematic
relationship. Looking at object type
and the temporal sequence with that objects are drawn, we can observe that
objects with line characteristics are frequently drawn first, most likely
because they can be used to define the drawing space and serve as local
reference frame. Annotations and direction indications are normally drawn at
the end of a sketch. In general it seems that a typical sketch has three
distinct phases that can be described as follows: The initial phase is
characterized by defining the rough framework of a sketch, the second phase
is used to populate the sketch with additional substantial objects, and the
last phase, finally, is responsible for finalizing details and the message of
a sketch. Although there is no iron rule for the composition of a sketch, we
observed that many sketches are created according to this concept. Based on our
questionnaire we have learned that there is almost no consent about what our
subjects considered to be difficult and what they think are easy sketching
tasks. Conversely, the approval of an appropriate size for an electronic
sketching device was much more evident. The averaged format comes close to
the US letter format (11 x 8.5") - despite 60% of our subjects
originating in Europe. Finally, we found that people would prefer for the
most part to interact also verbally with a computer system. Depending on the
complexity of the sketching task between 53% and 79% of our subjects stated
that they would like to provide verbal in addition to the graphical input.
This supports the idea that the future of computing belongs to multi-modal
systems. |
||
|
[Home] [Concept] [Potential
Applications] [Spatial
Theory] [User Interface] |
||
|
||
