Wolfgang Thaller

Irregular lattices for complex shape grammar facade parsing

High-quality urban reconstruction requires more than multi-view reconstruction and local optimization. The structure of facades depends on the general layout, which has to be optimized globally. Shape grammars are an established method to express hierarchical spatial relationships, and are therefore suited as representing constraints for semantic facade interpretation. Usually inference uses numerical approximations, or hard-coded grammar schemes. Existing methods inspired by classical grammar parsing are not applicable on real-world images due to their prohibitively high complexity. This work provides feasible generic facade reconstruction by combining low-level classifiers with mid-level object detectors to infer an irregular lattice. The irregular lattice preserves the logical structure of the facade while reducing the search space to a manageable size. We introduce a novel method for handling symmetry and repetition within the generic grammar. We show competitive results on two datasets, namely the Paris 2010 and the Graz 50. The former includes only Hausmannian, while the latter includes Classicism, Biedermeier, Historicism, Art Nouveau and post-modern architectural styles.

SMI 2013: Shape grammars on convex polyhedra

Shape grammars are the method of choice for procedural modeling of architecture. State of the art shape grammar systems define a bounding box for each shape; various operations can then be applied based on this bounding box. Most notably, the box can be split into smaller boxes along any of its three axes. We argue that a greater variety can be obtained by using convex polyhedra as bounding volumes instead. Split operations on convex polyhedra are no longer limited to the three principal axes but can use arbitrary planes. Such splits permit a volumetric decomposition into convex elements; As convex polyhedra can represent many shapes more faithfully than boxes, shape grammar rules can adapt to a much wider array of different contexts. We generalize established shape operations and introduce new operations that now become possible.

Procedural architecture using deformation-aware split grammars

With the current state of video games growing in scale, manual content creation may no longer be feasible in the future. Split grammars are a promising technology for large-scale procedural generation of urban structures, which are very common in video games. Buildings with curved parts, however, can currently only be approximated by static pre-modelled assets, and rules apply only to planar surface parts. We present an extension to split grammar systems that allow the creation of curved architecture through integration of free-form deformations at any level in a grammar. Further split rules can then proceed in two different ways. They can either adapt to these deformations so that repetitions can adjust to more or less space, while maintaining length constraints, or they can split the deformed geometry with straight planes to introduce straight structures on deformed geometry.

Deformation-Aware Split Grammars for Architectural Models

With the current state of video games growing in scale, manual content creation may no longer be feasible in the future. Split grammars are a promising technology for large scale procedural generation of urban structures, which are very common in video games. Buildings with curved parts, however, can currently only be approximated by static pre-modeled assets, and rules apply only to planar surface parts. We present an extension to current split grammar systems that allows the generation of curved architecture through free-form deformations that can be introduced at any level in a grammar. Further subdivision rules can then adapt to these deformations to maintain length constraints, and repetitions can adjust to more or less space.

Authoring Animated Interactive 3D Museum Exhibits using a Digital Repository

We present the prototype of a software system to streamline the serial production of simple interactive 3D animations for the display in museum exhibitions. We propose dividing the authoring process in two phases, a designer phase and a curator phase. The designer creates a set of configurable 3D scene templates that fit with the look of the physical exhibition while the curator inserts 3D models and configures the scene templates; the finished scenes are uploaded to 3D kiosks in the museum. Distinguishing features of our system are the tight integration with an asset repository and the simplified scene graph authoring. We demonstrate the usefulness with a few examples.

Procedural mesh features applied to subdivision surfaces using graph grammars

A typical industrial design modelling scenario involves defining the overall shape of a product followed by adding detail features. Procedural features are well-established in computer aided design (CAD) involving regular forms, but are less applicable to free-form modelling involving subdivision surfaces. Current approaches do not generate sparse subdivision control meshes as output, which is why free-form features are manually modelled into subdivision control meshes by domain experts. Domain experts change the local topology of the subdivision control mesh to incorporate features into the surface, without increasing the mesh density unnecessarily and carefully avoiding the appearance of artefacts.In this paper we show how to translate this expert knowledge to grammar rules. The rules may then be invoked in an interactive system to automatically apply features to subdivision surfaces. Graphical abstractDisplay Omitted HighlightsA novel method for adding freeform detail features to subdivision surfaces.Both input and output are industry-standard Catmull-Clark subdivision meshes.Expert knowledge about mesh topology modification is encoded as graph grammar rules.Resulting mesh contains few new control points than with comparable methods.Rules can encode domain-specific interactions between features.

Personalization of Virtual Coaching Applications using Procedural Modeling

Virtual coaching is an application area that allows individuals to improve existing skills or learn new ones; it ranges from simple textual tutoring tools to fully immersive 3D learning situations. The latter aim at improving the learning experience with realistic 3D environments. In highly individual training scenarios it can be beneficial to provide some level of personalization of the environment. This can be supported using procedural modeling that allows to easily modify shape, look and contents of an environment. We present the application of personalization using procedural modeling in learning applications in the project V2me. This project combines virtual and social networks to help senior citizens maintain and create meaningful relationships. We present a system that uses a procedurally generated ambient virtual coaching environment that can be adjusted by training subjects themselves or in collaboration. A small user experience study has been executed that gives first insight to the acceptance of such an approach.