![](bg3.png)
tents, metaphoric room connectors and camera view-
points marking points of interest.
The generative modeling paradigm, or procedu-
ral modeling, refers to parameters or lists of opera-
tions to generate 3D meshes and content using algo-
rithms rather than modeling interactively with CAD
software and exporting 3D models. Unlike the ap-
proach proposed in GML (Generative Modeling Lan-
guage) (TU Graz, 2010), our parameter design refers
not to low-level abstractions of components to de-
scribe 3D shapes such as points, lines or polygons,
but to an abstract representation of stylistic devices
such as (partition-)walls, windows, doorways or the
complete room shape. Parameters are specified in
the descriptive metadata sets of ViMCOX XML in-
stances, which are used by Replicave as configuration
for 3D content generation. Architectural design can
be specified as 2D floorplan in the ViMCOX meta-
data set. The 2D data will be transformed to 3D when
the virtual museum is generated. Floorplan data can
be acquired using different techniques. Modeling ap-
proaches based on 2D floorplans–using shape gram-
mar extensions and ontologies with export for dif-
ferent renderer (Tomas Trescak, 2012), transforma-
tion of Industry Foundation Classes (IFC) of build-
ings into CityGML XML instances as explicit geom-
etry (J. Benner, A. Geiger, K. Leinemann, 2005) or
2D floorplanning using graph theory (Ancona et al.,
2007)–can be found in the literature. Our concept
involves two means of generating content—metadata
and 3D content generation. This facilitates the ex-
change of authoring tools, the use of different 3D
visualisation techniques based on client capabilities
and dynamic generation of exhibitions based on larger
collections. Our metadata processor, the MetaToolkit
(Biella et al., 2012), provides interfaces and model
classes for automatic distribution of exhibits, addi-
tion of predefined viewpoints for each exhibit and
spatial modeling using algorithms. The MetaToolkit
uses JAXB (Java Architecture for XML Binding) to
accomplish a three tier schema mapping XML ↔
POJO → X3D. Schema-derived data classes gener-
ated by JAXB are transformed into corresponding
model classes to achieve convenient use in partial ap-
plications of ViMEDEAS and also to serve as input
for eased mapping of metadata to actual 3D content,
generated by Replicave (Biella et al., 2012).
An abstract architectural overview is presented in
Figure 1. Multimedia content; exported cultural ob-
ject metadata (LIDO); 3D models like (interactive)
exhibits, 3D buildings or 3D interior rooms; 2D pic-
tures; or floorplans as 2D point set can be used in
ViMCOX. Algorithms can facilitate 2D floorplanning
and automatic determination of exhibit distribution or
room layout. Authoring tools can modify the content
base as well as ViMCOX metadata instances. The dis-
semination and visualization layers are middleware
for assembling and publishing virtual museum in-
stances, locally or on the web, as well as interpreting
and rendering ViMCOX metadata instances on dif-
ferent visualization platforms. ViMEDEAS currently
only support X3D and X3Dom. Our generative mod-
eling approach becomes more efficient when larger
content bases need to be processed, which will be de-
scribed in the following application example as part of
our ongoing research (cf. section 3.1). The generated
show rooms will serve as an initial starting point for
further manual processing and visual design. Current
research addresses the development of API interfaces
to provide support for automatized generation of vir-
tual exhibition rooms based on simplified XML con-
figuration files and virtual objects stored as ViMCOX
instances. The exploratory stage focuses on paintings
and basic room shapes—rectangular shapes (aligned
to compass-directions) or room shapes in form of reg-
ular polygons. Our algorithm requires a list of walls
for each room and an ordered list of exhibitsand doors
to be placed on the corresponding walls. Parameters
for defining left and right spacing between objects are
optional and were initially set to 1.0m. The same
applies for heights of walls and vertical positioning
of the individual virtual objects. Initial textures for
walls, ceiling and floor are set to the Replicave pre-
set. The algorithm then calculates the individual wall
lengths and determines two side lengths (North-South
and West-East) for the construction of rectangular
rooms. For rooms in shape of regular polygons, the
largest wall will be used as the construction parame-
ter for calculating all wall coordinates. These calcula-
tions are done to preserve symmetric room shapes and
to determine additional spacing parameters to center
the objects as groups on walls, if desired. To calcu-
late wall lengths, the algorithm refers to the virtual
object dimension stored in the linked ViMCOX meta-
data instances. In the case of the ViMCOX object
type painting, the corresponding dimensions are the
image area, frame size and passe-partout size. Af-
terwards, the algorithm determines the positions of
each virtual painting or door and calculates a default
viewpoint for each object. The result is a single as-
sembled ViMCOX metadata instance, containing the
generated room instances with viewpoints, thresholds
(doors) and referenced objects as well as copies of the
initial virtual object instances. The generated virtual
rooms can then be visualized, locally or online, using
the Replicave framework (cf. section 3.2).
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
276