There is no data like more data -- current status of machine learning datasets in remote sensing

TL;DR

This paper reviews the evolution, features, and future challenges of machine learning datasets in remote sensing, emphasizing their importance for developing and evaluating automated interpretation methods.

Contribution

It provides a comprehensive overview of the history, current status, and open issues of remote sensing datasets for machine learning applications.

Findings

Datasets are crucial for remote sensing ML development.

Current datasets have limitations in diversity and scale.

Future work should focus on dataset standardization and expansion.

Abstract

Annotated datasets have become one of the most crucial preconditions for the development and evaluation of machine learning-based methods designed for the automated interpretation of remote sensing data. In this paper, we review the historic development of such datasets, discuss their features based on a few selected examples, and address open issues for future developments.