The Novel End-to-End Style Generation Model for Architectural Style Transfer and Classification

Architectural design is a multifaceted process that traditionally requires extensive time and expertise, often leading to inefficiencies and prolonged project timelines. This study explores the application of Artificial Intelligence (AI), specifically an optimized CycleGAN, to address these challenges by integrating architectural style transfer and classification into a single, end-to-end model. The model is trained on a diverse dataset comprising modern and traditional architectural styles, enabling it to transform modern architectural images into traditional styles while simultaneously categorizing the architectural style of each input. The results indicate that the model effectively generates traditional architectural styles from modern inputs, capturing key stylistic elements such as texture and form. However, the classification module reveals an imbalance in performance, with higher accuracy and lower loss for traditional style compared to modern style, suggesting areas for improvement. The findings underscore the potential of AI in revolutionizing architectural design, offering a more efficient, accurate, and streamlined process that can significantly enhance the precision and speed of architectural projects. This study contributes to the growing body of research on AI in architecture, demonstrating its applicability in real-world design scenarios.