SUGAMAN: Describing Floor Plans for Visually Impaired by Annotation Learning and Proximity based Grammar

In this paper, we propose SUGAMAN (Supervised and Unified framework using\nGrammar and Annotation Model for Access and Navigation). SUGAMAN is a Hindi\nword meaning "easy passage from one place to another". SUGAMAN synthesizes\ntextual description from a given floor plan image for the visually impaired. A\nvisually impaired person can navigate in an indoor environment using the\ntextual description generated by SUGAMAN. With the help of a text reader\nsoftware, the target user can understand the rooms within the building and\narrangement of furniture to navigate. SUGAMAN is the first framework for\ndescribing a floor plan and giving direction for obstacle-free movement within\na building. We learn $5$ classes of room categories from $1355$ room image\nsamples under a supervised learning paradigm. These learned annotations are fed\ninto a description synthesis framework to yield a holistic description of a\nfloor plan image. We demonstrate the performance of various supervised\nclassifiers on room learning. We also provide a comparative analysis of system\ngenerated and human written descriptions. SUGAMAN gives state of the art\nperformance on challenging, real-world floor plan images. This work can be\napplied to areas like understanding floor plans of historical monuments,\nstability analysis of buildings, and retrieval.\n