On the Status of Foundation Models for SAR Imagery

2025-09-25

AI Summary

Vision foundation models exhibit insufficient semantic feature extraction capability for synthetic aperture radar (SAR) image target recognition due to domain mismatch. Method: This paper proposes a novel paradigm of constructing SAR-specific foundation models via self-supervised fine-tuning. We systematically evaluate the adaptability of DINOv2, DINOv3, and PE-Core on SAR data and introduce the AFRL-DINOv2 series—domain-adapted variants trained via self-supervised fine-tuning to enhance feature discriminability. Contribution/Results: Extensive experiments demonstrate that AFRL-DINOv2 achieves superior robustness under low-label regimes and complex SAR scenes, consistently outperforming the state-of-the-art SARATR-X across multiple downstream tasks—including classification, detection, and segmentation—and establishing new performance records for SAR foundation models. The resulting architecture offers strong transferability and generalization, providing a scalable, task-agnostic backbone for intelligent SAR image analysis.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In this work we investigate the viability of foundational AI/ML models for Synthetic Aperture Radar (SAR) object recognition tasks. We are inspired by the tremendous progress being made in the wider community, particularly in the natural image domain where frontier labs are training huge models on web-scale datasets with unprecedented computing budgets. It has become clear that these models, often trained with Self-Supervised Learning (SSL), will transform how we develop AI/ML solutions for object recognition tasks - they can be adapted downstream with very limited labeled data, they are more robust to many forms of distribution shift, and their features are highly transferable out-of-the-box. For these reasons and more, we are motivated to apply this technology to the SAR domain. In our experiments we first run tests with today's most powerful visual foundational models, including DINOv2, DINOv3 and PE-Core and observe their shortcomings at extracting semantically-interesting discriminative SAR target features when used off-the-shelf. We then show that Self-Supervised finetuning of publicly available SSL models with SAR data is a viable path forward by training several AFRL-DINOv2s and setting a new state-of-the-art for SAR foundation models, significantly outperforming today's best SAR-domain model SARATR-X. Our experiments further analyze the performance trade-off of using different backbones with different downstream task-adaptation recipes, and we monitor each model's ability to overcome challenges within the downstream environments (e.g., extended operating conditions and low amounts of labeled data). We hope this work will inform and inspire future SAR foundation model builders, because despite our positive results, we still have a long way to go.

Problem

• Developing foundational AI models for SAR object recognition tasks
• Addressing shortcomings of existing visual models on SAR imagery
• Improving SAR model performance with limited labeled data

Innovation

• Self-supervised fine-tuning of SSL models
• Training AFRL-DINOv2 models with SAR data
• Analyzing backbone and adaptation recipe trade-offs