Enhancing Foundation Models for Imbalanced SAR Ship Classification via Targeted Oversampling

Ch Muhammad Awais; Marco Reggiannini; Davide Moroni

cs.CVSep 15, 2026

Enhancing Foundation Models for Imbalanced SAR Ship Classification via Targeted Oversampling

Authors: Ch Muhammad Awais, Marco Reggiannini, Davide Moroni

Organizations: Institute of Science and Technology, National Research Council, Pisa, Italy

Abstract

Remote-sensing foundation models offer strong representations for SAR imagery, but their behavior under severe long-tail class imbalance is still not well characterized. We benchmark DOFA and SAR-JEPA on the imbalanced OpenSARShip dataset and compare them with ImageNet-pretrained baselines under a fixed, training-efficient protocol that keeps the backbone frozen. To mitigate imbalance without fine-tuning, we apply four oversampling methods in embedding space exclusively to minority classes and train a lightweight classifier head on the augmented embeddings. Across both foundation models, oversampling improves Macro-F1 and test accuracy relative to their respective baselines, with the largest Macro-F1 gains observed for DOFA using ADASYN (34.39 to 38.56) and for SAR-JEPA using SVM-SMOTE (25.89 to 32.30). We also report class-wise behavior, showing that aggregate improvements can coexist with persistent failures on specific rare classes. Code for embedding extraction and reproducible multi-seed evaluation is provided to support rapid experimentation on free-tier hardware.

Figures & tables

Figure 1: Overview of the proposed evaluation pipeline. Given OpenSARShip, we extract frozen embeddings from each backbone (ImageNet baselines and Earth Observation foundation models), apply minority-only oversampling in embedding space, and train a lightweight fully connected classifier head on the augmented embeddings.

Model	Method	Accuracy	Macro-F1
DOFA	Base	60.65 $\pm$ 1.12	34.39 $\pm$ 0.26
DOFA	ADASYN	72.42 $\pm$ 0.84	38.56 $\pm$ 1.85
DOFA	KMeans	72.82 $\pm$ 0.75	35.59 $\pm$ 0.62
DOFA	SENN	66.96 $\pm$ 1.17	33.85 $\pm$ 1.23
DOFA	SVM	72.32 $\pm$ 1.36	36.87 $\pm$ 1.05
SAR-JEPA	Base	48.02 $\pm$ 2.53	25.89 $\pm$ 1.20

Table 1: Overall performance metrics.

DOFA
Class	Baseline	ADASYN	KMeans	SENN	SVM
Cargo	73.15 $\pm$ 1.27	82.52 $\pm$ 0.67	83.09 $\pm$ 0.64	81.32 $\pm$ 0.34	83.54 $\pm$ 0.19
Dredging	18.74 $\pm$ 3.64	19.93 $\pm$ 5.09	8.28 $\pm$ 4.46	19.71 $\pm$ 1.14	8.99 $\pm$ 0.14
Fishing	27.47 $\pm$ 2.05	24.70 $\pm$ 5.47	30.56 $\pm$ 2.82	32.03 $\pm$ 6.72	34.31 $\pm$ 4.52
Passenger	24.98 $\pm$ 7.47	38.41 $\pm$ 7.17	28.83 $\pm$ 3.88	26.08 $\pm$ 2.40	30.79 $\pm$ 1.12
Tanker	52.40 $\pm$ 1.39	51.30 $\pm$ 0.75	49.91 $\pm$ 1.64	30.75 $\pm$ 2.51	49.22 $\pm$ 0.63

Table 2: Per-class F1-scores across configurations.

Class-Imbalanced Learning

Class-imbalanced learning develops classifiers that recognize rare classes without sacrificing performance on common ones.

See all 256 papers

Cross-modal learning for SAR target recognition using optical vision foundation models
Sep 7, 2026Lucas Hirsch, James R. Hopgood, Javid Khan +2Cross-Modal Alignment Cross-Modal Representation Transfer
School of Engineering, University of Edinburgh, United Kingdom · Leonardo UK, United Kingdom · School of Engineering and Physical Sciences, Heriot-Watt University, United Kingdom
Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification
Sep 15, 2026Ch Muhammad Awais, Marco Reggiannini, Davide Moroni +1Classification Cross-Dataset Generalization
Institute of Science and Technology, National Research Council, Pisa, Italy
Prompting Foundation Models for Zero-Shot Ship Instance Segmentation in SAR Imagery
Apr 20, 2026Islam Mansour, Francescopaolo Sica, Michael SchmittInstance Segmentation Promptable Image Segmentation
Department of Aerospace Engineering, University of the Bundeswehr Munich, Neubiberg, Germany

Sep 7, 2026J/K move · Enter open · S save

Cross-modal learning for SAR target recognition using optical vision foundation models

Lucas Hirsch, James R. Hopgood, Javid Khan, Yoann Altmann+1

School of Engineering, University of Edinburgh, United Kingdom · Leonardo UK, United Kingdom · School of Engineering and Physical Sciences, Heriot-Watt University, United Kingdom

Cross-Modal Alignment Cross-Modal Representation Transfer Transfer Learning Synthetic Aperture Radar SAR ATR

Open paper AskPDF arXiv

Synthetic Aperture Radar (SAR) is an important modality in a wide range of imaging applications due to its versatile, long range and near all weather operating capabilities. However, Automatic Target Recognition (ATR) remains a challenging problem due to limited labelled data, the strong speckle in SAR images and the significant domain gap between SAR and more abundant optical imagery. In contrast, electro-optical (EO) imagery benefits from massive datasets, clearer visual structure and powerful foundation models. In this work, we investigate how vision foundation models trained on optical data can provide class level supervision for SAR classification. We propose a cross-modal EO to SAR prototype alignment framework in which a frozen EO encoder, based on a DINOv3 vision foundation model, is used to construct class level optical prototypes without requiring strict EO/SAR pairs. A SAR model is then trained to classify SAR images while aligning its embeddings to the corresponding EO class prototype. At inference time, the SAR model operates independently, without access to optical imagery. We evaluate our approach on the UNICORNv2 dataset, an EO and SAR dataset of civilian vehicles with heavily speckled images and severe class imbalance. EO prototype alignment improves SAR classification accuracy over frozen DINOv3, SAR only finetuning and unpaired distribution alignment baselines, and t-SNE visualizations provide qualitative evidence of clearer separation among classes in the trained SAR embedding space. These results suggest that optical vision foundation models, despite being trained on visible spectrum imagery, provide transferable information for SAR image classification, offering a practical method for using large scale pretrained vision foundation models across challenging sensing modalities.

Abstract

Class-Imbalanced Learning

Explore similar work

Cross-modal learning for SAR target recognition using optical vision foundation models

Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification

Prompting Foundation Models for Zero-Shot Ship Instance Segmentation in SAR Imagery