Abstract

Estimating 3D layout of a room from image data is a fundamental indoor scene understanding problem and has been a core research topic for over a decade. It is a necessary step for robot navigation, 3D reconstruction, and virtual, and augmented reality applications. Initially the problem was formulated to predict the layout from a single perspective image. Gradually research for this problem has expanded to estimating room layout from single panorama image, multiple images, and even from additional depth and range data. The estimated layout is usually represented as a set of projected corner positions, boundaries, or a 3D mesh. This problem has been approached many different ways over the years. This talk provides a brief overview of the existing research for solving the problem of room layout estimation.

About the Speaker

Mahbuba Perveen is a current PhD student at FloridaTech under the guidance of Dr. Eraldo Ribeiro. She received her BSc in Computer Science and Engineering from Jagannath University, Dhaka, Bangladesh. She is also one of the Teaching Assistants of the Computer Science department at Florida Tech.