Building Footprint Extraction from Remotely Sensed Data

The primary objective of this project is to automatically extract accurate footprint maps of buildings from remotely sensed data
such a satellite/aerial images. In recent times, deep learning approaches have shown significant progress and success in
processing remote-sensing image data. Most prevailing techniques used for this task often involve using pixel-level classification
deep learning techniques that involve extensive post-processing to generate the required building footprints. These methods often
fail to accurately capture the boundaries and corners of building footprints which are crucial in several downstream tasks.
Therefore, this project will focus on developing an end-to-end deep learning neural network architecture that can directly predict
accurate building footprint polygons from remote sensing data. These resulting building footprint polygons can be used directly in
downstream GIS, mapping, and reconstruction tasks, without the need for any post-processing operations.

Faculty Supervisor:

Charalambos Poullis

Student:

Partner:

Cyprus University of Technology

Discipline:

Computer science

Sector:

Artificial Intelligence; Construction

University:

Concordia University

Program: