A computational approach to understand building floor plan images using machine learning techniques

In recent past, there is a steep increase in the use of online platforms for the search for desired products. Real estate industry is no exception and has started initiating rent/sale of houses through online platforms. In this chapter, we present state of the art methods proposed in literature for automatic recognition and understanding of indoor spaces through their floor plan and RGB images and results. This chapter focuses on understanding of indoor spaces by analyzing their floor plans and generating a textual narration out of them. Floor plan image understanding is performed in a multistaged manner by recognizing visual components and generating a semistructured description for the floor plan image. Novel feature encoding is proposed for representing floor plan images for developing a machine learning based room recognition model. The quality of the generated narration is evaluated by comparing with the human written descriptions using state of the art language evaluation metrics. The proposed framework gives good quality descriptions of floor plans and has application in various applications such as, web-based real estate market, indoor navigation of robot or visual impaired person, similar plan retrieval.