Improving Architect-Specific Building Image Generation using Reinforced Data Processing

This paper investigates the innovative potential of fine-tuning generative Artificial Intelligence (AI) models through Reinforced Data Processing (RDP) for architectural visualization, focusing on the styles of three renowned architects. The study examines how the varying levels of detail in textual descriptions within image-text pairs impact model performance and evaluates these effects using a domain-specific RDP approach. The findings demonstrate that utilizing RDP models outperforms the base model in generating architect-specific images, offering a practical and creative tool for architects. This study underscores the potential of integrating generative AI into architectural design through RDP, offering substantial support for creative work.