Date of Award

May 2021

Degree Type


Degree Name

Master of Science



First Advisor

Istvan Lauko

Committee Members

David Spade, Dexuan Xi


Image to Image Translation, Neural Networks, Ultrasound, Vegetation Classification


Convolutional Neural Networks have been applied in many image applications, for both supervised and unsupervised learning. They have shown their ability to be used in an array of diverse use cases which include but are not limited to image classification, segmentation, and image enhancement tasks. We make use of Convolutional Neural Networks' ability to perform well in these situations and propose an architecture for a Convolutional Neural Network based on a network known as U-Net. We then apply our proposed network to two different tasks, a vegetation classification task for images of outdoors environment, and an image to image translation task for ultrasound images. For the vegetation classification task we make use of our previous work of a green vegetation filter that is used to annotate our data set and then use images that are converted to gray scale to pair with the annotations from the green vegetation filter in order to train our proposed network to classify where generic vegetation appears in an image. For the ultrasound image to image translation task, we show that our proposed network can be used as part of a system which is composed of a set of neural networks, called CycleGAN, that is used to translate ultrasound images from images acquired by a low frequency transducer to an image domain of ultrasound images acquired by a high frequency transducer. We propose using an approach that trains our proposed network to learn local estimations of the two image domains and detail a filtering process that when applied to an ultrasound image, acquired from a low frequency transducer, gives the low frequency transducer ultrasound image the appearance that it was acquired from a high frequency transducer.