Unsupervised Domain Adaptation for Land Cover Mapping

Overview

This research enhances land cover mapping using satellite imagery through unsupervised domain adaptation, addressing the challenge of limited labeled data in target regions.

Original research can be found here

Problem Statement

Approach

1. Dataset

2. Model Architecture

3. Unsupervised Domain Adaptation

4. Dynamic Pseudo-Labeling

5. Key Improvements

Results

Goal

Develop an effective, adaptable method for land cover mapping across different geographical regions with limited labeled data.

The project’s code and detailed results are available on GitHub.

View Project