SIGGRAPH 2003: Modeling and Simplification papers

This paper session seemed to be a bit of a catch-all, covering everything from a new UI for editing 3D objects to machine-created architecture and then finally the simulation of complex environments with texture depth maps.

Twister: A Space Warp Operator for the Two-Handed Editing of 3D Objects

Twister is a two-handed object manipulation device developed at Georgia Tech. It's intended to be used to manipulate all kinds of objects, although most of the examples in the talk involved extruding spheres or cubes in object creation. The idea has some promise, but is still mostly useful for course manipulation, as there is no fine selection method.

Instant Architecture

Although pretty mundane, this paper is a well-thought-out approach for dealing with the creation of architectural areas (cities, rural areas, etc.) by using a set of architectural data created by real humans who know what they are doing. Instead of allowing the computer to construct building without regard to architectural and design principles, this method provides a set of human- created rules for how building should look. These rules are then executed in order to create a building. Further, through integration with a GIS system (yeah, I know, I just said the equivalent of "ATM Machine"), areas can be designated as having certain kinds of buildings and the system can construct a viable cityscape.

The approach involves using hierarchical subdivision of the building area and the evaluation of the ruleset with a per-building set of random results in successive detail.

In the end, it is the grammar and the data represented therein that gives this technique its power, but the results are stunning. Efforts continue to create a larger database of rules that some day may be able to reasonably simulate an entire town or city.

Simulating Complex Environments Using Incremental Textured Depth Meshes

This technique extends the use of Textured Depth Meshes (TDMs) by incrementally adding more information to existing meshes based on a rendering samples made at different view locations.

TDMs are used to replace complex far-field data with textures that have depth and can reasonably simulate the visuals of distant objects. This is particularly useful in models having millions of polygons (or, as is the case in some real-world CAD systems, over a billion).

The technique attempts to fill in the gaps and stretches that are normally visible artifacts of using TDMs by using a Voronoi inspired set of sample locations to see into the void space (space about which nothing is known visually) from previous TDMs. However, instead of a normal TDM method, which would store another complete mesh, this method incrementally stores just the new data, thus allowing for a significant compression of the stored data.

On a sad note, the speaker dedicated his talk to his cat who died on Tuesday.