The goal of this independent study is to understand the impact of using superpixels on image segmentation tasks in deep learning. Image segmentation is the process of partitioning an image into distinct and meaningful regions. Superpixels are a group of pixels that share common characteristics such as color similarity and proximity. This paper will test the hypothesis that providing the U-Net with not only the raw image but also some additional context provided by an encoding of its superpixelation will help improve the U-Net's accuracy in segmenting the image.