Do Satellite Tasks Need Special Pretraining?

2025-10-19

AI Summary

This work investigates whether domain-specific pre-trained vision models are necessary for remote sensing—particularly under low-data and low-resolution conditions. Method: Leveraging the MillionAID dataset, the authors train a ViT-B encoder using the iBOT self-supervised framework, incorporating remote-sensing–specific architectural modifications and training strategies. They further construct a lightweight, multi-task remote sensing benchmark to systematically evaluate generalization across diverse downstream tasks. Contribution/Results: Empirical evaluation reveals that current remote-sensing–specific pre-training methods do not consistently outperform general-purpose foundation models (e.g., DINOv2, MAE) on mainstream downstream tasks—challenging the widely held assumption that domain specialization inherently yields superior performance. To our knowledge, this is the first study to empirically assess the necessity of remote-sensing–specific foundation models under a unified experimental protocol. The findings provide critical guidance for model selection in resource-constrained remote sensing applications.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: Other Foundations and Applications

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Foundation models have advanced machine learning across various modalities, including images. Recently multiple teams trained foundation models specialized for remote sensing applications. This line of research is motivated by the distinct characteristics of remote sensing imagery, specific applications and types of robustness useful for satellite image analysis. In this work we systematically challenge the idea that specific foundation models are more useful than general-purpose vision foundation models, at least in the small scale. First, we design a simple benchmark that measures generalization of remote sensing models towards images with lower resolution for two downstream tasks. Second, we train iBOT, a self-supervised vision encoder, on MillionAID, an ImageNet-scale satellite imagery dataset, with several modifications specific to remote sensing. We show that none of those pretrained models bring consistent improvements upon general-purpose baselines at the ViT-B scale.

Problem

• Evaluating specialized pretraining for satellite imagery analysis
• Comparing remote sensing foundation models with general vision models
• Testing model generalization on lower resolution satellite images

Innovation

• Self-supervised vision encoder trained on satellite dataset
• Modified pretraining approach for remote sensing imagery
• Benchmark measuring generalization to lower resolution images