Space Vector Generation for 3D Scenes from Text Descriptions

Abstract

3D scene generation using voice and text is convenient and low cost of learning. In the previous text-to-3D scene system, the user must use a fixed format for text input. In this paper, we present a newly system that generates space vector for 3D scenes from text natural language input. It is intended to benefit 3D scene producer and applications by generates space vector from text descriptions. This system consists of three parts: text processing, computational spatial relationships, and spatial vector generation. It computes the space vector by processing the text and then using the neural network to find the spatial relationships in the statement. The space vector is used to represent the spatial relationship between different objects in the text. This system eliminates the need for the user to control 3D scenes in fixed format text from the text to the 3D scene. The system is mainly used to process Chinese text, which can be used as a reference when dealing with other languages.