A study on a geometry transformation method for a geometry and scene modeler by a verbal interface

Abstract

The authors have developed a modeling system Do in three dimensional computer graphics (3D-CG) that, by verbal instructions in natural language, models geometrical shapes and renders the images. The former system is based on a step-by-step construction scheme using static instructions and a static display, that is, a static verbal instruction creates a static scene and an image. This staticbased system scheme has a critical limitation at the conversation stage between a user and the system. In this paper, the authors propose a dynamic instruction and dynamic display methods for the modeling system. First, the dynamic display enables users to watch the animation indicating the change between the previous geometries and the next geometries with smooth transition during the execution of an instruction from the user. Second, the dynamic instruction enables users to give the next instruction before the animation ends. For those purposes, two geometrical transformation methods by affine transformation: (a) with a common difference matrix, and (b) with a common ratio matrix, are introduced and studied to generate frames for the animation. The proposed methods are implemented and evaluated by some experiments. The changes between instructions are observed naturally by the common ratio matrix method.

An Adjectival Interface for Procedural Content Generation

“…Natural language has been used in conjunction with computer graphics in two major areas: in describing the relationships between objects [Kahn, 1979;Coyne and Sproat, 2001;Yanai and Okada, 2006] and in the tagging or labelling of objects [Barnard and Forsyth, 2001;Joshi et al, 2006]. Kahn [1979] represents some of the earliest work available, and describes the synthesis of animation from a high-level description of a film.…”

“…The major limiting factor of the system is the database of models available, although the user is free to expand this by adding their own models and associated tags. Yanai and Okada [2006] attack a different problem, namely the modelling of objects comprising simple geometric shapes. They describe a system that accepts natural language as input, and interprets this as a sequence of instructions for the step-by-step construction of a scene.…”

“…(a) A list of instructions to the system (b) The resulting object -a snowman Figure 28: An example of the technique of Yanai and Okada [2006], illustrating the construction of a snowman.…”