Convolutional neural network (CNN) can extract effective semantic features, so it was widely used for remote sensing image change detection (CD) in the latest years. CNN has acquired great achievements field of CD, but due to intrinsic locality convolution operation, could not capture global information space-time. The transformer proposed recent years and effectively information, solve computer vision (CV) tasks achieved amazing success. In this article, we design a pure with Siamese...

As an important task in the field of remote sensing (RS) image processing, RS change detection (CD) has made significant advances through use convolutional neural networks (CNNs). The transformer recently been introduced into CD due to its excellent global perception capabilities. Some works have attempted combine CNN and jointly harvest local-global features; however, these not paid much attention interaction between features extracted by both. Also, resulted resource consumption. In this...

Presently, research on deep learning-based change detection (CD) methods has become a hot topic. In particular, feature pyramid networks (FPNs) are widely used in CD tasks to gradually fuse semantic features. However, existing FPN-based do not correctly detect the complete region and cannot accurately locate boundaries of region. To solve these problems, new Multi-Scale Feature Progressive Fusion Network (MFPF-Net) is proposed, which consists three innovative modules: Layer Module (LFFM),...

In modern remote sensing image change detection (CD), convolution Neural Network (CNN), especially U-shaped structure (UNet), has achieved great success due to powerful discriminative ability. However, UNet-based CNN networks usually have limitations in modeling global dependencies intrinsic locality of operations. Transformer recently emerged as an alternative architecture for dense prediction tasks self-attention mechanism. limitation hardware resources, pure methods generally lack the...

In recent years, advancements in remote-sensing image super-resolution have achieved remarkable performance. However, many methods demand significant computational resources. This is problematic for edge devices with limited capabilities. To alleviate this problem, we propose an attention-based multi-level feature fusion network (AMFFN) to enhance the resolution of images. proposed integrates three efficient design strategies provide a lightweight solution. Initially, partial shallow...

Remote sensing image change detection (RSCD) is an important task in remote interpretation. Some recent RSCD works focus on the extraction and interaction of global local information. However, current work underutilizes hierarchical features may introduce noise from shallow encoders. In this paper, we propose a multi-scale cascaded cross-attention network (MSCCA-Net). This utilizes large kernel convolution formed by stacking small convolutions combined with Efficient Transformer as backbone...

Image features occur at different scales, e.g., short-term and long-term ones, both of them are significant in the change detection (CD) remote sensing images. To best our knowledge, however, it is still a challenge on how to effectively combine together for full-scale CD. The development deep learning techniques brings light this issue. In work, we propose hybrid initiative called HCGNet, combining convolutional neural network (CNN) vision graph (ViG) capturing local global features,...

The semantic segmentation of high-resolution remote sensing images (HRRSIs) faces persistent challenges in handling complex architectural structures and shadow occlusions, limiting the effectiveness existing deep learning approaches. To address these limitations, we propose an attention-focused feature enhancement network (AFENet) with a novel encoder–decoder architecture. encoder architecture combines ResNet50 parallel multistage group (PMFEG), enabling robust extraction through optimized...

As the field of remote sensing images processing continues to advance, semantic segmentation has become a focal point in this domain. The emergence Swin Transformer greatly alleviated computational complexities associated with Transformers, leading its widespread application segmentation. However, most current network models lack feature enhancement process internally, and model's tail lacks refinement modules prevent category misjudgments caused by redundancy. To address issue, we propose...

Deep hashing is the mainstream algorithm for large-scale cross-modal retrieval due to its high speed and low storage capacity, but problem of reconstruction modal semantic information still very challenging. In order further solve unsupervised reconstruction, we propose a novel deep semantic-preserving (DSPRH). The combines spatial channel information, mines based on adaptive self-encoding joint loss. main contributions are as follows: (1) We introduce new pooling network module tensor...

Change detection (CD) is one of the essential tasks in remote sensing image processing and analysis. Remote CD a process determining evaluating changes various surface objects over time. The impressive achievements deep learning computer vision provide an innovative concept for task CD. However, existing methods based on still have problems detecting small changed regions correctly distinguishing boundaries regions. To solve above shortcomings improve efficiency networks, inspired by fact...

Abstract The existing typical combined query image retrieval methods adopt Euclidean distance as sample measurement method, and the model trained by triple loss function blindly pursues absolute between samples, resulting in unsatisfactory performance. Meanwhile, these singularly Convolutional Neural Network (CNN) to extract reference features. However, receptive field of convolution operation has characteristics locality, which is easy cause edge feature information images. In view...

Recently, deep learning to hash has extensively been applied image retrieval, due its low storage cost and fast query speed. However, there is a defect of insufficiency imbalance when existing hashing methods utilize the convolutional neural network (CNN) extract semantic features extracted do not include contextual information lack relevance among features. Furthermore, process relaxation code can lead an inevitable quantization error. In order solve these problems, this paper proposes with...

Cross-modal retrieval is a very challenging and significant task in intelligent understanding. Researchers have tried to capture modal semantic information through weighted attention mechanism. Still, they cannot eliminate irrelevant information's negative effects fine-grained information. In order further accurately the multi-modal information, bidirectional focused alignment network (BFSAAN) proposed handle cross-modal tasks. Core ideas of BFSAAN are as follows: 1) Bidirectional mechanism...