抄録
A 3D virtual space creating system capable of customizing and learning is proposed in this paper. In the proposed system, a user inputs space parameters through a simple interface and virtual 3D spaces are generated. In order to obtain satisfactory spaces, a partial modification of spaces by user is possible. Furthermore, the system has the learning ability to support making spaces. When a user inputs the image of the generated space, the system learns the relation between various space parameters and the image by clustering and weighting of the parameters. After input of an image of 3D space that the user wants to make, the system generates the 3D space corresponding to the image. In this way, the early parameter settings are reduced in the proposed system. The followings have been confirmed by experiments : users can obtain satisfactory spaces by customization ; users can make spaces more easily by learning ability.
