Date of Award

May 2020

Degree Type

Thesis

Degree Name

Master of Science

Department

Computer Science

First Advisor

Zeyun Yu

Committee Members

Mukul Goyal, Sandeep Gopalakrishnan

Keywords

Deep Learning, Object Detection, Wound Localization, YOLOv3

Abstract

Developing a deep neural network for wound localization was the first step towards an efficient and fully automated wound healing system. A wound localizer was developed in this research using the YOLOv3 model, and an iOS mobile app was also created with the developed localization algorithm. The developed system can detect the wound and its surrounding tissue and isolate the portion of the localized wound for future care. This will support the segmentation and classification of wound by eliminating a lot of redundant details from photos of wound. A lighter variant of YOLOv3 called tiny-YOLOv3 is used for mobile device video processing. The model is trained and tested on an independently created dataset, designed in collaboration with AZH Wound and Vascular Center, Milwaukee, Wisconsin. Model YOLOv3 is contrasted with model SSD, showing that YOLOv3 gives 93.9% of the mAP value, which is much better than the SSD model (86.4%). These models’ robustness and reliability are shown to be very good when evaluated on a dataset that is publicly available.

Available for download on Wednesday, June 02, 2021

Share

COinS