Bridging Construction and AI: Computer Vision–Based Floor Plan Understanding for Intelligent Planning

Timely and accurate cost estimation is a critical determinant in fast-paced engineering construction projects. Traditional methods are often time-consuming and error-prone, relying on human interpretation of architectural drawings. With computer vision, the ability to provide early, reliable quotes can be accelerated, enabling fast, secure construction projects. In this study, we propose an automated computer vision framework that uses computer vision models to identify and quantify building components directly from floor and unit plans. The resulting database can be used to calculate the cost of detected components. Unlike existing methods that require engineers to interpret drawings, our approach establishes an end-to-end pipeline from drawing recognition to cost estimation, minimizing human effort and turnaround time. The proposed pipeline can process large construction drawing sets, automatically identify and extract relevant sheets such as floor plans, enlargement drawings, and schedules, and detect key building components like doors, windows, walls, and floors. Results show that model performance remains robust under different rotation rates, maintaining detection accuracy of 93%. Beyond immediate recognition of building components, this pipeline lays the groundwork for broader applications, such as material calculations and integration with building information modeling. In conclusion, the proposed method allows contractors to streamline human interpretation, secure a competitive advantage, and motivate the development of an AI-driven construction industry.