Advancement: Semantic Interior Mapology: An End-to-end Systematic Tool for Representing the Spatial Structure of an Indoor Environment

Speaker Name: 
Viet Trinh
Speaker Title: 
PhD Student (Advisor: Roberto Manduchi)
Speaker Organization: 
Computer Engineering
Start Time: 
Thursday, May 16, 2019 - 11:00am
End Time: 
Thursday, May 16, 2019 - 1:00pm
Engineering 2, Room 215
Roberto Manduchi

Abstract:  Due to the lack of visual cues and access to detailed environmental information, many visually impaired people are reluctant to travel independently in unfamiliar locations. Several lines of research have addressed the problem of wayfinding for blind individuals, mainly focusing on localization and guidance systems. Even when these systems are available, pre-journey learning is a valuable resource, helping blind people create a mental image of their surrounding and maintain their orientation in the case of system malfunctioning. Hence, in this dissertation, I propose Semantic Interior Mapology toolbox for the conversion of an architectural floor plan and its 3D-scanned room contents into a vectorized form that is amenable to reproduction in multiple modalities, serving the purpose of pre-journey learning. The toolbox allows one to quickly trace the layout of a building to produce a spatial features representation. Objects appearing in the 3D scans of rooms within the building are embedded automatically as small-scale features. In addition, the toolbox also generates tactile maps at different levels of details on demand, providing blind travelers with a spatial understanding of an indoor environment. Semantic Interior Mapology minimizes the time and effort required to acquire a detailed description of an indoor space, and produces accurate results even in the case of complex building layouts.