🤖 AI Summary
Machine learning research in Earth observation has long been hindered by benchmark datasets that are disconnected from real-world applications and exhibit saturated performance, resulting in poor model generalizability and deployment reliability.
Method: This paper proposes a data-centric paradigm tailored to geospatial data, systematically formalizing its methodological framework and clarifying its complementary relationship with the prevailing model-centric paradigm. Emphasizing data quality, representativeness, and closed-loop iteration, it integrates key techniques—including data cleaning, annotation optimization, synthetic augmentation, distribution alignment, active learning, and feedback-driven refinement—into an end-to-end pipeline spanning problem formulation, data optimization, modeling, deployment, and feedback integration.
Contribution/Results: Experiments demonstrate that this paradigm significantly improves model accuracy and robustness on unseen scenarios, breaks through benchmark performance ceilings, and substantially enhances practical applicability in operational Earth observation systems.
📝 Abstract
Recent developments and research in modern machine learning have led to substantial improvements in the geospatial field. Although numerous deep learning architectures and models have been proposed, the majority of them have been solely developed on benchmark datasets that lack strong real-world relevance. Furthermore, the performance of many methods has already saturated on these datasets. We argue that a shift from a model-centric view to a complementary data-centric perspective is necessary for further improvements in accuracy, generalization ability, and real impact on enduser applications. Furthermore, considering the entire machine learning cycle — from problem definition to model deployment with feedback — is crucial for enhancing machine learning models that can be reliable in unforeseen situations. This work presents a definition as well as a precise categorization and overview of automated data-centric learning approaches for geospatial data. It highlights the complementary role of data-centric learning with respect to model-centric in the larger machine learning deployment cycle. We review papers across the entire geospatial field and categorize them into different groups. A set of representative experiments shows concrete implementation examples. These examples provide concrete steps to act on geospatial data with data-centric machine learning approaches.
