Deep Learning Approach to Estimate the Optimal Number of Piles and Beams from Architectural Floorplans

The traditional process of estimating piles and beams from architectural floor plans is a manual, time-consuming task that is prone to human error. Accurate estimation is vital for effective planning and execution of construction projects. It helps in scheduling, resource allocation, and logistics, ensuring that the construction process is smooth and efficient. The need for an automated process to enhance efficiency and accuracy is a primary problem being addressed. This paper presents an innovative framework for automating the estimation of the optimal number of piles and beams from architectural floor plans using advanced image processing and structural analysis techniques. Leveraging a deep segmentation network with a context-aware classifier, we enhance the accuracy of identifying key structural elements in floor plans. Our method involves a three-stage process: extracting structural elements, vectorizing floor plans, and identifying load-bearing walls. We employ a thinning algorithm and contour reduction techniques for precise vectorization, and our approach in determining room spans assists in accurately locating load-bearing walls. This method-ology not only streamlines the estimation process of foundational elements but also introduces a novel way of integrating deep learning with architectural engineering, setting a new standard in construction planning. Preliminary results demonstrate a significant reduction in time and error margins compared to traditional methods, showcasing the potential of our framework to revolutionize construction planning and execution.