Leveraging Machine Learning Techniques in Converting 2D Floorplans Images to 3D Models: Applications, Challenges, and Future Directions

Abstract

With the growing demand for automation in fields such as architecture, real estate, and digital twin technologies, the ability to efficiently convert 2D floorplan images into accurate 3D structural models has become increasingly critical. Traditional CAD (Computer-Aided Design)-based approaches, while precise, often lack scalability and adaptability in dynamic or large-scale environments. In response, recent advancements in machine learning have opened new possibilities for intelligent 3D reconstruction. This short review explores these developments, surveying key machine learning pipelines, benchmark datasets, evaluation metrics, and real-world applications. It also addresses persistent challenges including generalization, occlusion, and dataset limitations. The paper highlights promising directions such as diffusion models and foundation models, that aim to overcome current barriers and shape the future of automated 3D modeling from 2D sources. This work contributes to a clearer understanding of the current landscape, identifies gaps in existing approaches, and outlines strategic pathways for future research in data-driven architectural modeling.