Get our free extension to see links to code for papers anywhere online!

Chrome logo  Add to Chrome

Firefox logo Add to Firefox


Polygonal Building Segmentation by Frame Field Learning

Apr 30, 2020
Nicolas Girard, Dmitriy Smirnov, Justin Solomon, Yuliya Tarabalka


Share this with someone who'll enjoy it:


While state of the art image segmentation models typically output segmentations in raster format, applications in geographic information systems often require vector polygons. We propose adding a frame field output to a deep image segmentation model for extracting buildings from remote sensing images. This improves segmentation quality and provides structural information, facilitating more accurate polygonization. To this end, we train a deep neural network, which aligns a predicted frame field to ground truth contour data. In addition to increasing performance by leveraging multi-task learning, our method produces more regular segmentations. We also introduce a new polygonization algorithm, which is guided by the frame field corresponding to the raster segmentation.



   Access Paper Source



Share this with someone who'll enjoy it: