Abstract
Space planning and interior design require not only technical precision but also creative thinking and spatial awareness. Although earlier research has examined the cognitive and educational elements that influence spatial ability, such as fuzzy DEMATEL and ISM-based models, these studies lack real-time decision-making support, machine-aided creativity, and practical implementation. To overcome these limitations, this study suggests an intelligent framework for enhancing interior design and space planning. The originality of this work lies in bridging spatial ability cognitive insights and machine interaction from a human-centered perspective to real-world design applications and an AI-powered design process. The proposed work combines hybrid models, including CenterNet and StyleGAN3, along with a transformer model. In order to read user intent or constraints (text, sketch, or spatial signals), the suggested system first uses CenterNet to detect and map spatial aspects. StyleGAN3 relies on these semantic and spatial features to generate interior visuals that are congruent with style. Finally, the transformer model is used as a sequence-aware logic engine responsive to user interaction. As opposed to existing work that focuses mainly on instructional modeling, this work presents machine-aided co-design with enhanced utility and beauty as well as user preference consideration. Categorized user intents, room layout, and high-resolution interior style images were some of the data applied to the experimental evaluation. While StyleGAN3 generated high-quality and varied outputs (FID = 11.6), the suggested system used CenterNet to achieve excellent results in spatially accurate localization (mAP@0.75 = 84.3%). Human assessments with artists revealed increased usability satisfaction and support for creativity (avg. score: 4.5/5). This demonstrates the effectiveness of integrating generative artistic cognition with organized spatial thinking within a collaborative AI system.