An Imbalanced Discriminant Alignment Approach for Domain Adaptive SAR Ship Detection

Shi, Zhenwei; Xu, Zhehao; Shi, Tianyang; Li, Tao; Shi, Zhenwei

doi:10.1109/tgrs.2023.3303507

Cited by 8 publications

(1 citation statement)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Shi et al [23] extended the pioneering work of domain-adapted object detection by integrating multiple discriminators by Chen et al [24]. Pan et al [25] proposed an end-to-end domain-adaptation-based ship detection network consisting of imbalanced discriminant alignment and imbalanced prediction consistency. Xu et al [26] introduced a multilevel alignment network that transfers knowledge from the optical domain to the SAR domain using domain adversarial strategies.…”

Section: Introductionmentioning

confidence: 99%

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

Luo,

Zhang,

Guo

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

The unique imaging modality of synthetic aperture radar (SAR) has posed significant challenges for object detection, making it more complex to acquire and interpret than optical images. Recently, numerous studies have proposed cross-domain adaptive methods based on convolutional neural networks (CNNs) to promote SAR object detection using optical data. However, existing cross-domain methods focus on image features, lack improvement on input data, and ignore the valuable supervision provided by few labeled SAR images. Therefore, we propose a semi-supervised cross-domain object detection framework that uses optical data and few SAR data to achieve knowledge transfer for SAR object detection. Our method focuses on the data processing aspects to gradually reduce the domain shift at the image, instance, and feature levels. First, we propose a data augmentation method of image mixing and instance swapping to generate a mixed domain that is more similar to the SAR domain. This method fully utilizes few SAR annotation information to reduce domain shift at image and instance levels. Second, at the feature level, we propose an adaptive optimization strategy to filter out mixed domain samples that significantly deviate from the SAR feature distribution to train feature extractor. In addition, we employ Vision Transformer (ViT) as feature extractor to handle the global feature extraction of mixed images. We propose a detection head based on normalized Wasserstein distance (NWD) to enhance objects with smaller effective regions in SAR images. The effectiveness of our proposed method is evaluated on public SAR ship and oil tank datasets.

show abstract

Section: Introductionmentioning

confidence: 99%

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

Luo,

Zhang,

Guo

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

show abstract

Adaptive Multiscale Reversible Column Network for SAR Ship Detection

Wang,

Zeng

2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Ship detection via Synthetic Aperture Radar(SAR) is widely used in maritime safety and maritime traffic control, et. al. Recently, deep learning methods are employed to improve the performance of SAR ship detection to a large extent. However, the presence of clutter in SAR images and the large scale difference of ships result in the diminished detection performance in complex environments. As such, in this paper, a novel Adaptive Multi-Scale Reversible Column Network is proposed. Firstly, the idea of disentangled feature learning is applied to construct Reversible Column Networks with C2f module to alleviate the problem of large scale difference and the loss of ship information. Secondly, the multiplexed adaptive spatial pyramid pooling module is proposed to alleviate the impact of complex background clutter through multiplexed pooling operations and adaptive fusion. Finally, a novel feature pyramid network with adaptive downsampling module is designed to reduce the information loss caused by downsampling while enhancing the multi-scale ship detection capability. The effectiveness of the proposed method is validated on two public datasets: SSDD and HRSID. The experimental results show that the proposed method is ablet o achieve better results than current state-of-the-art methods for SAR ship detection in complex environments and large scale difference of ships.

show abstract

CroMoDa: Unsupervised Oriented SAR Ship Detection via Cross-Modality Distribution Alignment

Chen,

Wang,

Wang

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

An Imbalanced Discriminant Alignment Approach for Domain Adaptive SAR Ship Detection

Cited by 8 publications

References 41 publications

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

Adaptive Multiscale Reversible Column Network for SAR Ship Detection

CroMoDa: Unsupervised Oriented SAR Ship Detection via Cross-Modality Distribution Alignment

Contact Info

Product

Resources

About