Towards Scalable and Generalizable Earth Observation Data Mining via Foundation Model Composition
This work addresses the need for scalable and generalizable solutions in Earth Observation data mining, offering a practical approach for real-world applications, though it is incremental as it builds on existing models and ensembling techniques.
The study tackled the problem of improving Earth Observation data mining by investigating whether combining existing pretrained foundation models could enhance performance across diverse tasks, and found that feature-level ensembling of smaller models matched or exceeded larger models while reducing training time and computational resources.
Foundation models are rapidly transforming Earth Observation data mining by enabling generalizable and scalable solutions for key tasks such as scene classification and semantic segmentation. While most efforts in the geospatial domain have focused on developing large models trained from scratch using massive Earth Observation datasets, an alternative strategy that remains underexplored is the reuse and combination of existing pretrained models. In this study, we investigate whether foundation models pretrained on remote sensing and general vision datasets can be effectively combined to improve performance across a diverse set of key Earth Observation tasks. Using the GEO-Bench benchmark, we evaluate several prominent models, including Prithvi, Hiera, and DOFA, on eleven datasets covering a range of spatial resolutions, sensor modalities, and task types. The results show that feature-level ensembling of smaller pretrained models can match or exceed the performance of much larger models, while requiring less training time and computational resources. Moreover, the study highlights the potential of applying knowledge distillation to transfer the strengths of ensembles into more compact models, offering a practical path for deploying foundation models in real-world Earth Observation applications.