Computational Imagination in Three-Dimension Scene Synthesis
Chong Yu Zeng, Yue Wu, Zisen Zhou, Yanju Yao, Hongji Yang · 2024
Integrating emotional and behavioural data into 3D scene synthesis is an emerging area blending cognitive science with computational creativity to create more engaging, realistic, and emotionally resonant virtual environments. This study presents an interdisciplinary framework for generating 3D scenes from textual narratives, where emotional and physical metaphors guide the rendering process. The system utilises natural language processing techniques to identify behavioural patterns and emotional values in user-supplied sentences. These emotions determine the absolute position of the subject in the spatial matrix, placing those subjects with stronger emotional content closer to the observer. At the same time, the relative positions of subjects, predicates and objects utilise the metaphor of gravity to simulate an attraction-rejection dynamic similar to that of physical bodies in space. This ingenious positioning system is based on the laws of physics and allows for the construction of more situational 3D environments. The proposed model demonstrates how computational imagination can leverage physical principles to significantly enhance creativity in visual narrative and interaction design.