Consistency-Regularized GAN for Few-Shot SAR Target Recognition

2026-01-22

IEEE Transactions on Geoscience and Remote Sensing

AI Summary

This work addresses the challenges of synthetic aperture radar (SAR) image target recognition under extremely limited sample availability and the instability of conventional GAN training by proposing a Consistency-Regularized Generative Adversarial Network (Cr-GAN). The method employs a dual-branch discriminator to decouple adversarial training from representation learning and integrates channel-level feature interpolation with dual-domain cycle-consistency regularization. This enables the generation of high-fidelity, diverse SAR images even under ultra-low-shot conditions, facilitating self-supervised pretraining and fine-tuning. Cr-GAN is compatible with various GAN architectures and self-supervised algorithms, achieving state-of-the-art accuracy of 71.21% on MSTAR and 51.64% on SRSDD under an 8-shot setting—significantly outperforming existing approaches—while using only approximately one-fifth the number of parameters of advanced diffusion models.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Few-shot recognition in synthetic aperture radar (SAR) imagery remains a critical bottleneck for real-world applications due to extreme data scarcity. A promising strategy involves synthesizing a large dataset with a generative adversarial network (GAN), pre-training a model via self-supervised learning (SSL), and then fine-tuning on the few labeled samples. However, this approach faces a fundamental paradox: conventional GANs themselves require abundant data for stable training, contradicting the premise of few-shot learning. To resolve this, we propose the consistency-regularized generative adversarial network (Cr-GAN), a novel framework designed to synthesize diverse, high-fidelity samples even when trained under these severe data limitations. Cr-GAN introduces a dual-branch discriminator that decouples adversarial training from representation learning. This architecture enables a channel-wise feature interpolation strategy to create novel latent features, complemented by a dual-domain cycle consistency mechanism that ensures semantic integrity. Our Cr-GAN framework is adaptable to various GAN architectures, and its synthesized data effectively boosts multiple SSL algorithms. Extensive experiments on the MSTAR and SRSDD datasets validate our approach, with Cr-GAN achieving a highly competitive accuracy of 71.21% and 51.64%, respectively, in the 8-shot setting, significantly outperforming leading baselines, while requiring only ~5 of the parameters of state-of-the-art diffusion models. Code is available at: https://github.com/yikuizhai/Cr-GAN.

Problem

• Few-shot learning
• SAR target recognition
• Data scarcity
• Generative adversarial network
• Synthetic aperture radar

Innovation

• Consistency-Regularized GAN
• Few-Shot SAR Recognition
• Dual-Branch Discriminator
• Channel-wise Feature Interpolation
• Dual-Domain Cycle Consistency