Storing the semantics embedded in digital shapes

Digital 3D shapes, whether modelling a human kidney or the door-handle on a concept car, are crucial to modern science and industrial practice. Currently in widespread use for computer-aided design and manufacture, they are becoming crucial to genomic, proteomic and medical modelling.

Digital shapes are used extensively to develop models and create simulations. Also to devise new designs that conform to engineering constraints, yet remain functional and aesthetically pleasing. In personal entertainment, it is already possible to buy your own 3D character for some computer games.

Yet research on digital shapes suffers strongly from a lack of the ability to re-use existing knowledge, 3D models and software. Current projects typically begin from scratch, developing everything they need by themselves. The AIM@SHAPE project partners hope to make that approach a thing of the past.

AIM@SHAPE is a Network of Excellence established in January 2004 and due to finish in December 2007. If the partners succeed in their aims, it will establish good practice for the use and re-use of digital shapes, which, especially in industrial design, could mean enormous savings in time, money and creative effort.

Describing the shape and nature of a model, however, is a serious challenge.This is equally true whether that shape is a complex composite design, or a simple element within the overall whole. The mathematical descriptions of a CAD model, a triangular mesh used in a computer game and medical raster data from an MRI scan, are all vastly different.

Another issue is how to describe the semantics, the knowledge connected to a shape. If only the shape's geometry is stored in an archive, the knowledge connected to it, like its history, its purpose and intent, will be lost. That is why the semantic information must remain attached to the object, and in a way that computing systems can create, recognise and manipulate automatically, without the need for human intervention such as entering hundreds of keywords.

AIM@SHAPE sought to develop a foundation, a theoretical model to illustrate how a shape can be described, not only in terms of a shape's quantitative geometric properties but also in semantic context, including software tools. This approach allowed the project team to develop integrated ontologies, semantic descriptions, as machine-readable code, both for the shapes and for the research software using these figures.

Armed with that foundation theory, the project developed specialised ontologies, or machine-readable dictionaries, for three application domains, relating to the fields of virtual human beings, product design, and shape acquisition and processing.

“These ontologies work well for the three areas, and they combine knowledge related to the general geometric domain with information specific to these areas. Moreover, if you wanted to create a scene where a human hand is grasping a chair, you could in principle combine the metadata stored by the virtual-human and product-design ontologies to fully describe the shapes, and use that information to categorise the file and find it later,” explains project coordinator Bianca Falcidieno.

The project team chose these three domains because they were of particular interest to the partners. With the methodology now established, it should be possible to use them as blueprints to create new ontologies or adapt existing ones to describe other domains, e.g. geographic information services, architecture and many more.

The project team has also developed an infrastructure called the 'Digital Shape Workbench', which integrates different repositories of resources for the storage, manipulation and management of shapes.

The 'shape repository' stores shapes with particular properties as a collection of benchmarks and standard cases, while the 'tool repository' contains state-of-the-art research software tools for processing shapes. The platform's 'digital library' integrates bibliographies, or book lists, of information either generated or recommended by members of the consortium. The 'ontology and metadata repository' integrates shapes and tools with semantic information, and allows the user to search and analyse the content of the workbench.

For the final phase of the project over the next year, the project partners are focusing on further developing this search engine, with a semantic component for reasoning and a geometry-based component for content-based 3D shape matching and retrieval.

However, the current software tools are highly sophisticated and require expert skills to use them. In the future of 3D shapes, she would therefore like to see the development of simpler tools that can be used by non-experts. Conceivably, she believes, people could one day manipulate 3D models or create their own digital designs with the same ease as they manipulate photos now.

Beyond the technical goals for the remainder of the project, however, the AIM@SHAPE partners want to continue their work. Digital shapes and particularly 3D design are likely to be huge fields, touching every domain of human activity. Already some applications are leaking into the mainstream market, for example interior-design programs for PCs, or video-game editors that allow players to design their own characters.

AIM@SHAPE is well positioned to help European companies gain a head start in this potentially huge market. Before that happens, however, the partners will need to establish the network on an institutional or associative basis, in order to continue the work for the long term, and to involve new partners from academia and industry for new shape-and-semantics research and development projects.