Floor Plan Designer Application by Predicting Spatial Configuration Using Machine Learning

Abstract

In the field of architecture, floor plan design is the way to generate various floor plans for houses. Designing of floor plan is one of the most essentials for all kinds of infrastructures. In spite of this, there are many computerized and manual ways are available today. Nowadays, the burden of manual work has been greatly reduced. Even though the manual works have been greatly reduced, still there are some manual works like dragging and dropping, aligning the rooms, placing in desired locations, etc. We cannot determine the exact lengths of rooms, hall, kitchen, etc. Floor plan designing is one of the highly iterative processes and it demands extensive human labour; however, they are limited to fully automate the creative process. This proposed work faces so many real challenges because of the many implicit and explicit rules in order to create viable floor plans. In this work, a floor plan designer is suggested for an automated floor plan generating application that can generate floor plans automatically when the total area of the land and choice of BHK (B = Bedroom. H = Hall. K = Kitchen) is provided by the customers. In this research work, deep convolutional neural network (DCNN) methods are proposed to handle floor plan parsing and spatial recognition methods. Spatial recognition [1] among objects plays a fundamental role in the human perception and understanding of design. The various spatial recognition methods are used to edit the generated floor plans based on their willingness and acceptance.