Nick Dunn

Hi, I’m Nick. I am a master’s student in computer science at Carnegie Mellon University (CMU). I earned my bachelor’s degrees in both computer science and mathematics in 2024 from Columbus State University, where I was also a tutor for both subjects.

During the summer of 2023, I worked as an undergraduate researcher in the Robotics Institute at CMU as part of the Robotics Institute Summer Scholars (RISS) program. My research focused on improving instance segmentation performance of computer vision machine learning models. I was fortunate to be advised by Dr. Christoph Mertz and collaborate with several others in the CMU community.

During the summer of 2021, I worked as an undergraduate researcher in the Department of Computer Science and Software Engineering at Auburn University as part of the Research Experiences for Undergraduates (REU) on Smart Unmanned Aerial Vehicles (UAVs) program funded by the National Science Foundation (NSF). Our research focused on autonomously navigating UAVs through a GPS-denied environment using computer vision, onboard cameras and sensors, and fiducial markers.

I also previously worked as a firefighter/EMT 🚒🔥🚑 and as a telecommunications technician providing phone 📞 and internet 💻 service to the military.

Selected Publications

GeometryPaste: Geometry-Based Copy-Paste Data Augmentation for Instance Segmentation

Nicholas Dunn, Anurag Ghosh, and Christoph Mertz

RISS Working Papers Journal, 2023

Abs PDF Poster

Instance segmentation models require large datasets of annotated images to achieve adequate performance. However, annotated datasets are difficult to build or obtain. There are several large-scale datasets for common objects, but few exist for rare objects. Detecting rare objects has many practical applications, such as autonomous vehicles detecting roadwork objects. Copy-Paste is a data augmentation method for generating images and has been utilized successfully to improve instance segmentation performance. Prior works have studied both random Copy-Paste, where objects are randomly pasted onto images, and pasting objects based on the surrounding visual context. In this paper, we develop GeometryPaste, a method of pasting objects according to the geometry and context of the objects and background images. We build a small dataset of roadwork objects and fine-tune a pre-trained instance segmentation model to evaluate our method. Our results are compared against both baseline and random Copy-Paste APs. The results suggest that GeometryPaste may provide performance improvements over both baseline and random Copy-Paste augmentation for instance segmentation of rare object categories in small datasets.
Autonomous UAV Navigation in a GPS-Denied Outdoor Environment Using Discontinuous Visual Contact with Fiducial Markers

Nicholas Dunn, Zachary Schallenberger, Saad Biaz, and 1 more author

J. Comput. Sci. Coll., 2022

Abs Poster

Autonomous navigation of Unmanned Aerial Vehicles (UAVs) has become an increasingly popular area of study as technology has advanced over the years. One technology that has been developed for use in the autonomous navigation of UAVs is fiducial markers. Fiducial markers are a type of physical tag that can be placed in an environment to assist with UAV localization, navigation, and landing. A UAV can identify and scan a fiducial marker with an onboard camera and take appropriate action based on the data received from the marker. Previous research using only fiducial markers and onboard sensors for UAV navigation in a GPS-denied environment requires the camera to have constant visual contact with at least one marker. This paper explores the use of fiducial markers for outdoor UAV navigation in a GPS-denied environment without the restriction of constant visual contact with a marker. Physical implementation and formal methods are used to study the performance and overall viability of this solution.