Advancements in Learning-Based Techniques for Automated Floor Plan Analysis on Raster Images: A Comparative Study

Abstract

Automated floor plan analysis and recognition have long been focal points in computer science research. Recently, there has been a notable increase in the use of learning-based techniques to automatically reorganize floor plans from raster images. This advancement aims to extract valuable insights from architectural drawings, which are essential for understanding building layouts and their intended functions. These drawings often feature a variety of notations and constraints, and the lack of standardized notation leads to significant variability in both style and semantics across different floor plans. Addressing this challenge is a key focus of this review. This paper provides an extensive literature survey to tackle the issue of variability in floor plans. The review concentrates on methodologies that treat floor plans as raster images, with particular attention to learning-based approaches. By offering concise summaries of datasets, research scopes, and specific tasks, this review aims to guide future research and development in the fields of construction and design. The in-depth examination of automatic floor plan analysis and recognition methods presented here contributes to the evolving field of computer-assisted architectural understanding and design.