Unsupervised image segmentation in satellite imagery using deep learning

Author

Suraj Regmi

Date of Award

2023

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Chair

Huaming Zhang

Committee Member

Manil Maskey

Committee Member

Chaity Banerjee Mukherjee

Subject(s)

Remote sensing, Image segmentation, Deep learning (Machine learning)

Abstract

Image segmentation is typically done through supervised learning. Supervised learning requires labeled data, which is costly and time-consuming to acquire. However, unlabeled data is abundant. This research presents an application of the state-of-the-art unsupervised instance segmentation method FreeSOLO in satellite images and benchmarks the method in iSAID, CrowdAI, and PASTIS datasets. The method achieved 0.9%AP₅₀ in the iSAID dataset, 3.1%AP₅₀ on the CrowdAI dataset, and 1.1%AP₅₀ on the PASTIS dataset. On large objects, it achieved 1.2%AP₅₀ in the iSAID dataset and 3.5%AP₅₀ in the CrowdAI dataset. The method was also tested in the UAH periphery and MBRSC Dubai dataset where the model was able to segment buildings, water bodies, highways, apartments, and trees. This research also demonstrates the comparative performance of FreeSOLO-based weights relative to other popular supervised learning-based encoder weights on semantic segmentation downstream task.

Recommended Citation

Regmi, Suraj, "Unsupervised image segmentation in satellite imagery using deep learning" (2023). Theses. 458.
https://louis.uah.edu/uah-theses/458