In this project we introduce low cost methods of improving dilated convolutions in an image segmentation application. We achieve comparable results to state-of-the-art segmentation performance while being computationally more efficient than previously proposed methods.