SAR Target Detection Based on SSD With Data Augmentation and Transfer Learning

Wang, Zhaocheng; Du, Lan; Mao, Jiashun; Liu, Bin; Yang, Dongwen

doi:10.1109/lgrs.2018.2867242

Cited by 126 publications

(65 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The hardware capabilities include NVIDIA RTX-2080 GPU (8GB memory), Intel® i7-8700 CPU @3.20GHz and 32 GB RAM. To maintain the same hyperparameters of the detectors, we choose mmdetection (a flexible toolkit for reimplementing existing methods) [ (2) 1000x1000 pixels in the process of training and testing [66][67][68][69]. All the detectors are trained with GPU and finished in 12th epochs; the momentum and weight decay are set to 0.9 and 0.0001, respectively.…”

Section: B Experimental Detailsmentioning

confidence: 99%

HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation

et al. 2020

View full text Add to dashboard Cite

With the development of satellite technology, up to date imaging mode of synthetic aperture radar (SAR) satellite can provide higher resolution SAR imageries, which benefits ship detection and instance segmentation. Meanwhile, object detectors based on convolutional neural network (CNN) show high performance on SAR ship detection even without land-ocean segmentation; but with respective shortcomings, such as the relatively small size of SAR images for ship detection, limited SAR training samples, and inappropriate annotations, in existing SAR ship datasets, related research is hampered. To promote the development of CNN based ship detection and instance segmentation, we have constructed a High-Resolution SAR Images Dataset (HRSID). In addition to object detection, instance segmentation can also be implemented on HRSID. As for dataset construction, under the overlapped ratio of 25%, 136 panoramic SAR imageries with ranging resolution from 1m to 5m are cropped to 800 x 800 pixels SAR images. To reduce wrong annotation and missing annotation, optical remote sensing imageries are applied to reduce the interferes from harbor constructions. There are 5604 cropped SAR images and 16951 ships in HRSID, and we have divided HRSID into a training set (65% SAR images) and test set (35% SAR images) with the format of Microsoft Common Objects in Context (MS COCO). 8 state-of-the-art detectors are experimented on HRSID to build the baseline; MS COCO evaluation metrics are applicated for comprehensive evaluation. Experimental results reveal that ship detection and instance segmentation can be well implemented on HRSID.

show abstract

Section: B Experimental Detailsmentioning

confidence: 99%

HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The recorder was used to store spatial features of labeled samples, and the recorded features were used to predict the labels of unlabeled data points based on spatial similarity to increase the number of labeled samples. Finally, Weng et al [13] used an approach more similar to our framework. Their proposal was to transfer knowledge from the EO domain using VGGNet as a feature extractor in the learning pipeline, which itself has been pretrained on a large EO dataset.…”

Section: Related Workmentioning

confidence: 99%

“…The general idea that we focus on is to transfer knowledge from a secondary domain to reduce the amount labeled data that are necessary to train a model. Building upon prior works in the area of transfer learning, several recent works have used the idea of knowledge transfer to address the challenges of SAR domains [6,8,10,[12][13][14]. The common idea in these works is to transfer knowledge from a secondary related problem, where labeled data are easy and inexpensive to obtain.…”

mentioning

confidence: 99%

Deep Transfer Learning for Few-Shot SAR Image Classification

et al. 2019

View full text Add to dashboard Cite

The reemergence of Deep Neural Networks (DNNs) has lead to high-performance supervised learning algorithms for the Electro-Optical (EO) domain classification and detection problems. This success is because generating huge labeled datasets has become possible using modern crowdsourcing labeling platforms such as Amazon’s Mechanical Turk that recruit ordinary people to label data. Unlike the EO domain, labeling the Synthetic Aperture Radar (SAR) domain data can be much more challenging, and for various reasons, using crowdsourcing platforms is not feasible for labeling the SAR domain data. As a result, training deep networks using supervised learning is more challenging in the SAR domain. In the paper, we present a new framework to train a deep neural network for classifying Synthetic Aperture Radar (SAR) images by eliminating the need for a huge labeled dataset. Our idea is based on transferring knowledge from a related EO domain problem, where labeled data are easy to obtain. We transfer knowledge from the EO domain through learning a shared invariant cross-domain embedding space that is also discriminative for classification. To this end, we train two deep encoders that are coupled through their last year to map data points from the EO and the SAR domains to the shared embedding space such that the distance between the distributions of the two domains is minimized in the latent embedding space. We use the Sliced Wasserstein Distance (SWD) to measure and minimize the distance between these two distributions and use a limited number of SAR label data points to match the distributions class-conditionally. As a result of this training procedure, a classifier trained from the embedding space to the label space using mostly the EO data would generalize well on the SAR domain. We provide a theoretical analysis to demonstrate why our approach is effective and validate our algorithm on the problem of ship classification in the SAR domain by comparing against several other competing learning approaches.

show abstract

“…YOLO and its variants [27]- [31] are known because of their fast speed and high efficiency. SSD and its variants [32]- [36] blend the advantages of these two methods. However, for moving object detection, these deep learning architectures face several critical issues:…”

Section: Related Workmentioning

confidence: 99%

Moving Object Detection With Deep CNNs

et al. 2020

View full text Add to dashboard Cite

In large field of view for open country, the real-time detection and identification of moving objects with high accuracy is a very challenging work due to the excessive amount of data. This paper proposes a novel framework that consists of a coarse-grained detection as well as a fine-grained detection. To solve the problem of noise-induced object fracture during the coarse-grained detection process, we present a low-complexity connected region detection algorithm to extract moving regions. Furthermore, in the finegrained detection, Deep Convolution Neural Networks are leveraged to detect more precise coordinates and identify the category of objects. To the best of our knowledge, this is the first work that proposes a coarse-tofine grained framework to detect moving objects on high-resolution scenes. Experimental results show that the proposed framework can robustly work on the high resolution video frames (1920*1080p) with complex situations more fastly and accurately over existing methods. INDEX TERMS Connected region detection, deep convolution neural networks, foreground extraction, high resolution, moving object detection.

show abstract

SAR Target Detection Based on SSD With Data Augmentation and Transfer Learning

Cited by 126 publications

References 11 publications

HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation

HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation

Deep Transfer Learning for Few-Shot SAR Image Classification

Moving Object Detection With Deep CNNs

Contact Info

Product

Resources

About