Abstract
In the study of Marangoni condensation through thermal images, a common task is to analyze the features of a single droplet. To do this, the droplet must first be segmented from the background. While this segmentation can be done by hand for small batch sizes, scalability quickly becomes an issue. As machine learning tools have historically been strong in image segmentation, this study analyzes the efficacy of implementing a U-Net machine learning model for this specific segmentation task. Through the process, tools were developed to help in the creation of training data, and algorithms were fine tuned to the specifics of the segmentation.
Document Type
Final Report
Class Name
Mechanical Engineering and Material Sciences Independent Study
Language
English (en)
Date of Submission
12-22-2025
Recommended Citation
Rubel, Brendan, "Investigating Droplet Segmentation With Machine Learning Tools" (2025). Mechanical Engineering and Materials Science Independent Study. 315.
https://openscholarship.wustl.edu/mems500/315