Search

Rotated object detection in remote sensing images presents a highly challenging task due to the extensive fields of view and complex backgrounds. While Convolutional Neural Networks (CNNs) and Transformer networks have made progress in this area, there is still a lack of research on extracting and fusing features for small targets in complex backgrounds. To address this gap, we have extended the RTMDet framework by introducing three modules: the Focused Feature Context Aggregation Module, the Feature Context Information Enhancement Module, and the Multi-scale Feature Fusion Module. In the Focused Feature Context Aggregation Module, we replaced the Spatial Pyramid Pooling Bottleneck (SPPFBottleneck) to better extract small target features by focusing on contextual information. The Feature Context Information Enhancement Module enhances the model’s perception of multi-dimensional temporal and spatial information. Finally, we combined the original features with the fused ones to prevent the loss of specific features during the fusion process. Our proposed model, named the Multi-scale Feature Context Aggregation Network (MFCANet), was evaluated on four challenging remote sensing datasets (MAR20, SRSDD, HRSC, and DIOR-R). The experimental results demonstrate that our method outperforms baseline models, achieving improvements of 2.13%, 10.28%, 1.46%, and 1.13% in mAP for the MAR20, SRSDD, HRSC, and DIOR-R datasets, respectively.

High-resolution multi-spectral (HRMS) images have been widely used in various fields, however, they can not be directly obtained due to the physical hardware limits of existing remote sensing satellites. Therefore, the pansharpening technique has been widely explored as an effective tool to generate HRMS images by fusing the complementary information of low-resolution multi-spectral (LRMS) images and high-resolution panchromatic (PAN) images. Existing deep learning-based pansharpening methods mainly focus on enhancing the spatial representation ability of the network, while paying little attention to modeling spectral dependencies in spite of its significance for remote sensing data. In this paper, we propose a simple yet effective channel group shuffling (CGS) module to explore the implicit relationships with regard to the adjacent and cross-channels while considering spatial information. To be specific, the proposed CGS module consists of two components: the channel group module and the feature shuffle fusion module. The former enhances the diversity of spectral information and cross-channel information communications while ensuring the spectral order of the input feature. The latter integrates the cross-group feature maps with rich spatial-spectral information. Equipped with the proposed functional module, our image fusion network, dubbed CGSNet, produces favorable results against existing state-of-the-art counterparts over various satellite datasets. Ablation studies further verify the flexibility and effectiveness of our core design.

With the improvement of quality of life, people are more and more concerned about the quality of sleep. The electroencephalogram (EEG)-based sleep stage classification is a good guide for sleep quality and sleep disorders. At this stage, most automatic staging neural networks are designed by human experts, and this process is time-consuming and laborious. In this paper, we propose a novel neural architecture search (NAS) framework based on bilevel optimization approximation for EEG-based sleep stage classification. The proposed NAS architecture mainly performs the architectural search through a bilevel optimization approximation, and the model is optimized by search space approximation and search space regularization with parameters shared among cells. Finally, we evaluated the performance of the model searched by NAS on the Sleep-EDF-20, Sleep-EDF-78 and SHHS datasets with an average accuracy of 82.7%, 80.0% and 81.9%, respectively. The experimental results show that the proposed NAS algorithm provides some reference for the subsequent automatic design of networks for sleep classification.

Rapid progress has been made in the research of high-resolution remote sensing road extraction tasks in the past years but due to the diversity of road types and the complexity of road context, extracting the perfect road network is still fraught with difficulties and challenges. Many convolutional neural networks (CNNs) based on encoder–decoder structures have demonstrated their effectiveness. Transformer’s self-attention mechanism shows more powerful performance than CNNs in modeling global feature dependencies. In this letter, we propose a multiscale deformable transformer network (MDTNet) based on encoder–decoder structure to extract road networks from remote sensing images. The core of MDTNet is our proposed multiscale deformable self-attention (MDSA) mechanism. MDSA can capture more comprehensive features than conventional self-attention. In addition, roads are not present in certain blocks of areas like other objects, but are interwoven throughout the image in such a long, linear fashion that information about certain road segments may be overlooked. To minimize residual errors in road segmentations, our MDSA incorporates a deformable design on feature maps, which effectively enhances the salience of road features relative to their surroundings. Extensive experiments on several public remote sensing road datasets show that our MDTNet achieves higher segmentation [F1 score and intersection over union (IoU)] and connectivity [average path length similarity (APLS)] accuracy, which verifies the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

In this article, we study two challenging problems in semisupervised cross-view learning. On the one hand, most existing methods assume that the samples in all views have a pairwise relationship, that is, it is necessary to capture or establish the correspondence of different views at the sample level. Such an assumption is easily isolated even in the semisupervised setting wherein only a few samples have labels that could be used to establish the correspondence. On the other hand, almost all existing multiview methods, including semisupervised ones, usually train a model using a fixed dataset, which cannot handle the data of increasing views. In practice, the view number will increase when new sensors are deployed. To address the above two challenges, we propose a novel method that employs multiple independent semisupervised view-specific networks (ISVNs) to learn representation for multiple views in a view-decoupling fashion. The advantages of our method are two-fold. Thanks to our specifically designed autoencoder and pseudolabel learning paradigm, our method shows an effective way to utilize both the labeled and unlabeled data while relaxing the data assumption of the pairwise relationship, that is, correspondence. Furthermore, with our view decoupling strategy, the proposed ISVNs could be separately trained, thus efficiently handling the data of increasing views without retraining the entire model. To the best of our knowledge, our ISVN could be one of the first attempts to make handling increasing views in the semisupervised setting possible, as well as an effective solution to the noncorresponding problem. To verify the effectiveness and efficiency of our method, we conduct comprehensive experiments by comparing 13 state-of-the-art approaches on four multiview datasets in terms of retrieval and classification. [ABSTRACT FROM AUTHOR]