Jacob Eggerd
Pronouns: He/Him
Research Mentor(s): Dawen Cai
Research Mentor School/College/Department: Cell and Developmental Biology; Biophysics; Neuroscience Graduate Program / Medicine
Program:
Authors: Jacob Eggerd, Logan Walker, Dawen Cai
Session: Session 2: 10:00 am – 10:50 am
Poster: 41
Abstract
When taking microscopic images, several different resolutions can be chosen. A higher resolution objective provides more detail while capturing proportionally less of the image. There are several machine learning algorithms designed to increase image resolution. Taking higher-resolution images that cover the same amount of the sample requires more labor. We will demonstrate that an SRGAN model generates satisfactory super-resolution images for samples. Our dataset was developed from TetraSpeck Microspheres at low and high resolutions. We then found an SRGAN model that was previously trained on nonmicroscopic images. Adding our microscopic dataset will hopefully lead it to adapt well to biomedical data. We hope to see the results demonstrate that an SRGAN model with our dataset is a viable option for upscaling these microscopic images. It can allow lower-resolution images to be captured without having to sacrifice the quality of them. Considering how fundamental our dataset is, we believe the model will generalize to many other applications. The TetraSpeck images are abstract and apply well to images of any kind. It should broaden the utility of the SRGAN network.
