Semester

Summer

Date of Graduation

2022

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Gianfranco Doretto

Committee Co-Chair

Donald Adjeroh

Committee Member

Donald Adjeroh

Committee Member

Omid Dehzangi

Abstract

Automated cellular instance segmentation is a process that has been utilized for accelerating biological research since before the deep learning era, and recent advancements have produced higher quality results with less effort from the biologist. Most current endeavors focus on completely cutting the researcher out of the picture by generating highly generalized models. However, these models invariably fail when faced with novel data and effectively opt to miss out on the full capabilities of deep learning in pursuit of this goal. In our work, we demonstrate how, with even a minimal amount of annotated data, dominant approaches in this space can achieve much better performance in many cases and cover blind spots that generalized models cannot accommodate. Utilizing specialized contrastive losses, our few-shot domain adaptation method not only performs well on all datasets tested; it even approaches the level of networks trained directly on the target datasets. In several ways, this work extends the current aims of generalized cellular instance segmentation techniques and paves a path toward an optimal balance between model performance and expert-level annotation.

Embargo Reason

Patent Pending

Share

COinS