Channel Dimensions Research Articles

MAFormer: A cross-channel spatio-temporal feature aggregation method for human action recognition

Human action recognition has been widely used in fields such as human–computer interaction and virtual reality. Despite significant progress, existing approaches still struggle with effectively integrating hierarchical information and processing data beyond a certain frame count. To address these challenges, we introduce the Multi-AxisFormer (MAFormer) model, which is organized in terms of spatial, temporal, and channel dimensions of the action sequence, thereby enhancing the model’s understanding of correlations and intricate structures among and within features. Drawing on the Transformer architecture, we propose the Cross-channel Spatio-temporal Aggregation (CSA) structure for more refined feature extraction and the Multi-Axis Attention (MAA) module for more comprehensive feature aggregation. Moreover, the integration of Rotary Position Embedding (RoPE) boosts the model’s extrapolation and generalization abilities. MAFormer surpasses the known state-of-the-art on multiple skeleton-based action recognition benchmarks with the accuracy of 93.2% on NTU RGB+D 60 cross-subject split, 89.9% on NTU RGB+D 120 cross-subject split, and 97.2% on N-UCLA, offering a novel paradigm for hierarchical modeling in human action recognition.

DRAINAGE SYSTEM TRANSFORMATION TOWARDS EFFECTIVE FLOOD CONTROL WITH EPA SWMM

The village of Kunjir often experiences floods caused by suboptimal drainage systems, land use changes, and revetment construction. The utilization of EPA SWMM proves to be an interesting solution to address flood issues. The research aims to analyze the drainage channel capacity using the EPA SWMM program, assess the tidal influence on the channels, and provide solutions to the existing problems. The method involves hydrological and hydraulic analysis using Microsoft Excel and the EPA SWMM program. Rainfall and topographic data are utilized to determine flood discharge, while existing channel data are used to calculate channel capacity. HWL values are used to analyze the tidal influence on the channels. The modeling process involves creating a network scheme and inputting data into the existing channels. The program is executed, and evaluations are conducted on problematic channels. Redesign planning is implemented by altering channel dimensions and re-running the simulations. The conclusion indicates that channels 2B, 2B1, 2B1 Left, 2C, and 2D cannot accommodate the flow discharge, tidal influence affects channel 1C, Box Culvert, and Downstream River, and solutions include enlarging channels, adding bottom slope to channels, regular maintenance, and installing floodgates to mitigate tidal effects.

Dual Channel‐Spatial Self‐Attention Transformer and CNN synergy network for 3D medical image segmentation

Even though the Vision Transformer leverages the self-attention mechanism to capture long-range dependencies, showing significant potential in medical image segmentation, the limited annotations in the image dataset make it difficult for the Transformer model to extract different global features, resulting in attention collapse and generating similar or identical attention maps. Previous studies have attempted to solve the problem by integrating convolutional neural layers into Transformer-based architectures. However, improper integration may lead to the inability of the model to effectively capture local and global information in both spatial and channel dimensions. To address the above issue, we propose a hybrid architecture using the Dual Channel-Spatial Self-Attention Transformer and CNN Synergy Network (DTC-SUNETR) for medical image segmentation. Specifically, we redesigned the self-attention mechanism. A novel Channel-Spatial Self-Attention (CSSA) block is introduced to integrate the enhanced channel and spatial self-attention mechanism to capture the global relationship and local structure among image features. This helps the model to more comprehensively understand the inter-dependencies between different channels and capture the relationships between different pixels, thus enhancing the feature representation of the corresponding dimensions. Simultaneously, it also improves the overall computational efficiency of the network. Extensive experiments on four different medical image segmentation datasets, including Synapse, ACDC, Brain Tumor, and Lung Tumor, demonstrate the superiority of the proposed DTC-SUNETR over state-of-the-art methods.

Automatic Water Body Extraction from SAR Images Based on MADF-Net

Water extraction from synthetic aperture radar (SAR) images has an important application value in wetland monitoring, flood monitoring, etc. However, it still faces the problems of low generalization, weak extraction ability of detailed information, and weak suppression of background noises. Therefore, a new framework, Multi-scale Attention Detailed Feature fusion Network (MADF-Net), is proposed in this paper. It comprises an encoder and a decoder. In the encoder, ResNet101 is used as a solid backbone network to capture four feature levels at different depths, and then the proposed Deep Pyramid Pool (DAPP) module is used to perform multi-scale pooling operations, which ensure that key water features can be captured even in complex backgrounds. In the decoder, a Channel Spatial Attention Module (CSAM) is proposed, which focuses on feature areas that are critical for the identification of water edges by fusing attention weights in channel and spatial dimensions. Finally, the high-level semantic information is effectively fused with the low-level edge features to achieve the final water detection results. In the experiment, Sentinel-1 SAR images of three scenes with different characteristics and scales of water body are used. The PA and IoU of water extraction by MADF-Net can reach 92.77% and 89.03%, respectively, which obviously outperform several other networks. MADF-Net carries out water extraction with high precision from SAR images with different backgrounds, which could also be used for the segmentation and classification of other tasks from SAR images.

Scalable video transformer for full-frame video prediction

Vision Transformers (ViTs) have shown success in many low-level computer vision tasks. However, existing ViT models are limited by their high computation and memory cost when generating high-resolution videos for tasks like video prediction. This paper presents a scalable video transformer for full-frame video predication. Specifically, we design a backbone transformer block for our video transformer. This transformer block decouples the temporal and channel features to reduce the computation cost when processing large-scale spatial–temporal video features. We use transposed attention to focus on the channel dimension instead of the spatial window to further reduce the computation cost. We also design a Global Shifted Multi-Dconv Head Transposed Attention module (GSMDTA) for our transformer block. This module is built upon two key ideas. First, we design a depth shift module to better incorporate the cross-channel or temporal information from video features. Second, we introduce a global query mechanism to capture global information to handle large motion for video prediction. This new transformer block enables our video transformer to predict a full frame from multiple past frames at the resolution of 1024 × 512 with 12 GB VRAM. Experiments on various video prediction benchmarks demonstrate that our method with only RGB input outperforms state-of-the-art methods that require additional data, like segmentation maps and optical flows. Our method exceeds the state-of-the-art RGB-only methods by a large margin (1.2 dB) in PSNR. Our method is also faster than state-of-the-art video prediction transformers.

DMMGNet: A discrimination mapping and memory bank mean guidance-based network for high-performance few-shot industrial anomaly detection

For deep learning-based industrial anomaly detection, it is still challenging to get adequate images for model training and achieve cold start for cross-product migration, restricting their practical application in real industrial production. Herein, an innovative few-shot anomaly detection network DMMGNet based on discrimination mapping and memory bank mean guidance strategies are demonstrated, which is trained by a new two-branch data augmentation technique. By separating the features stored in memory bank from the features used for training, the two-branch data augmentation method can significantly improve the robustness of few-shot model training and reduce the redundance of memory bank. In the elaborately designed discrimination mapping module, new negative samples are generated by adding dynamic Gaussian noise to normal samples along the channel dimension in feature space to solve the problem of sample imbalance. Meanwhile, the discrimination mapping module also helps to map the feature distribution of positive samples to the target domain more efficiently and reduce the deviation of feature domain, conducive to a more precise separation of positive and negative samples. In addition, a novel mean guidance approach with an optimized loss function is developed to guide the positive sample feature mapping by specifying the local feature space center to form a clear feature domain contour and enhance the detection accuracy. The multiple experimental results validate that our DMMGNet outperforms the most advanced anomaly detection counterparts on image-level AUROC, showing an increase by 0.3–3 % on both MVTec AD and MPDD benchmarks under several few-shot scenarios.

A One-Dimensional Computational Model to Identify Operating Conditions and Cathode Flow Channel Dimensions for a Proton Exchange Membrane Fuel Cell

A one-dimensional computational model has been developed that can be used to identify operating conditions for the cathode side of a proton exchange membrane fuel cell such that both the inlet and outlet relative humidity is equal to 100%. By balancing the calculated pressure drop along the cathode side flow channel with the change in molar composition, inlet conditions for the cathode side can be identified with the goal of avoiding channel flooding. The channel length, height, width and the land-to-channel width ratio are input parameters for the model so that it might be used to dimension the cathode flow field. The model can be used to calculate the limiting current density, and we are presenting unprecedented high values as a result of the high pressure drop along the flow channels. Such high current densities can ultimately result in a fuel cell power density beyond the typical value of 1.0–2.0 W/cm2 for automotive fuel cells.

Flexible droplet microfluidic devices for tuneable droplet generation

Droplet microfluidics is a promising technology for applications that require precise handling of minute fluid volumes. This technology has broad applications in the pharmaceutical industry, food and beverage, and material synthesis. Droplet size and frequency are the two key parameters in these applications. Current droplet microfluidic devices are rigid platforms where the geometry and dimensions of microchannels are fixed after the device fabrication. This poses a significant limitation when adjusting the droplet generation characteristics. Repetitive design, fabrication, and testing are often needed to achieve the optimal microchannel dimensions. To overcome this bottleneck, we develop a proof-of-concept stretchable microfluidic device that can control droplet size and generation frequency by precisely controlling the microchannel dimensions in real-time. Theoretical analysis, numerical modelling, and experimental characterisation were conducted to study the influence of device lateral stretching on these characteristics. We found that the lateral stretching of the device increased the droplet diameter and spacing but reduced the droplet generation frequency. Droplet diameter and spacing increased by ∼20 %, and droplet frequency decreased by ∼45 % when the device was strained up to 25 %. We believe this innovative, flexible droplet microfluidic platform will provide an alternative way to precisely control the droplet formation by modifying the channel dimensions on-site and in real-time.

Attention-Based Spatiotemporal-Aware Network for Fine-Grained Visual Recognition

On public benchmarks, current macro facial expression recognition technologies have achieved significant success. However, in real-life scenarios, individuals may attempt to conceal their true emotions. Conventional expression recognition often overlooks subtle facial changes, necessitating more fine-grained micro-expression recognition techniques. Different with prevalent facial expressions, weak intensity and short duration are the two main obstacles for perceiving and interpreting a micro-expression correctly. Meanwhile, correlations between pixels of visual data in spatial and channel dimensions are ignored in most existing methods. In this paper, we propose a novel network structure, the Attention-based Spatiotemporal-aware network (ASTNet), for micro-expression recognition. In ASTNet, we combine ResNet and ConvLSTM as a holistic framework (ResNet-ConvLSTM) to extract the spatial and temporal features simultaneously. Moreover, we innovatively integrate two level attention mechanisms, channel-level attention and spatial-level attention, into the ResNet-ConvLSTM. Channel-level attention is used to discriminate the importance of different channels because the contributions for the overall presentation of micro-expression vary between channels. Spatial-level attention is leveraged to dynamically estimate weights for different regions due to the diversity of regions’ reflections to micro-expression. Extensive experiments conducted on two benchmark datasets demonstrate that ASTNet achieves performance improvements of 4.25–16.02% and 0.79–12.93% over several state-of-the-art methods.

EFS-Former: An Efficient Network for Fruit Tree Leaf Disease Segmentation and Severity Assessment

Fruit is a major source of vitamins, minerals, and dietary fiber in people’s daily lives. Leaf diseases caused by climate change and other factors have significantly reduced fruit production. Deep learning methods for segmenting leaf diseases can effectively mitigate this issue. However, challenges such as leaf folding, jaggedness, and light shading make edge feature extraction difficult, affecting segmentation accuracy. To address these problems, this paper proposes a method based on EFS-Former. The expanded local detail (ELD) module extends the model’s receptive field by expanding the convolution, better handling fine spots and effectively reducing information loss. H-attention reduces computational redundancy by superimposing multi-layer convolutions, significantly improving feature filtering. The parallel fusion architecture effectively utilizes the different feature extraction intervals of the convolutional neural network (CNN) and Transformer encoders, achieving comprehensive feature extraction and effectively fusing detailed and semantic information in the channel and spatial dimensions within the feature fusion module (FFM). Experiments show that, compared to DeepLabV3+, this method achieves 10.78%, 9.51%, 0.72%, and 8.00% higher scores for mean intersection over union (mIoU), mean pixel accuracy (mPA), accuracy (Acc), and F_score, respectively, while having 1.78 M fewer total parameters and 0.32 G lower floating point operations per second (FLOPS). Additionally, it effectively calculates the ratio of leaf area occupied by spots. This method is also effective in calculating the disease period by analyzing the ratio of leaf area occupied by diseased spots. The method’s overall performance is evaluated using mIoU, mPA, Acc, and F_score metrics, achieving 88.60%, 93.49%, 98.60%, and 95.90%, respectively. In summary, this study offers an efficient and accurate method for fruit tree leaf spot segmentation, providing a solid foundation for the precise analysis of fruit tree leaves and spots, and supporting smart agriculture for precision pesticide spraying.

Tiny-Object Detection Based on Optimized YOLO-CSQ for Accurate Drone Detection in Wildfire Scenarios

Wildfires, which are distinguished by their destructive nature and challenging suppression, present a significant threat to ecological environments and socioeconomic systems. In order to address this issue, the development of efficient and accurate fire detection technologies for early warning and timely response is essential. This paper addresses the complexity of forest and mountain fire detection by proposing YOLO-CSQ, a drone-based fire detection method built upon an improved YOLOv8 algorithm. Firstly, we introduce the CBAM attention mechanism, which enhances the model’s multi-scale fire feature extraction capabilities by adaptively adjusting weights in both the channel and spatial dimensions of feature maps, thereby improving detection accuracy. Secondly, we propose an improved ShuffleNetV2 backbone network structure, which significantly reduces the model’s parameter count and computational complexity while maintaining feature extraction capabilities. This results in a more lightweight and efficient model. Thirdly, to address the challenges of varying fire scales and numerous weak emission targets in mountain fires, we propose a Quadrupled-ASFF detection head for weighted feature fusion. This enhances the model’s robustness in detecting targets of different scales. Finally, we introduce the WIoU loss function to replace the traditional CIoU object detection loss function, thereby enhancing the model’s localization accuracy. The experimental results demonstrate that the improved model achieves an mAP@50 of 96.87%, which is superior to the original YOLOV8, YOLOV9, and YOLOV10 by 10.9, 11.66, and 13.33 percentage points, respectively. Moreover, it exhibits significant advantages over other classic algorithms in key evaluation metrics such as precision, recall, and F1 score. These findings validate the effectiveness of the improved model in mountain fire detection scenarios, offering a novel solution for early warning and intelligent monitoring of mountain wildfires.

CasUNeXt: A Cascaded Transformer With Intra‐ and Inter‐Scale Information for Medical Image Segmentation

ABSTRACTDue to the Transformer's ability to capture long‐range dependencies through Self‐Attention, it has shown immense potential in medical image segmentation. However, it lacks the capability to model local relationships between pixels. Therefore, many previous approaches embedded the Transformer into the CNN encoder. However, current methods often fall short in modeling the relationships between multi‐scale features, specifically the spatial correspondence between features at different scales. This limitation can result in the ineffective capture of scale differences for each object and the loss of features for small targets. Furthermore, due to the high complexity of the Transformer, it is challenging to integrate local and global information within the same scale effectively. To address these limitations, we propose a novel backbone network called CasUNeXt, which features three appealing design elements: (1) We use the idea of cascade to redesign the way CNN and Transformer are combined to enhance modeling the unique interrelationships between multi‐scale information. (2) We design a Cascaded Scale‐wise Transformer Module capable of cross‐scale interactions. It not only strengthens feature extraction within a single scale but also models interactions between different scales. (3) We overhaul the multi‐head Channel Attention mechanism to enable it to model context information in feature maps from multiple perspectives within the channel dimension. These design features collectively enable CasUNeXt to better integrate local and global information and capture relationships between multi‐scale features, thereby improving the performance of medical image segmentation. Through experimental comparisons on various benchmark datasets, our CasUNeXt method exhibits outstanding performance in medical image segmentation tasks, surpassing the current state‐of‐the‐art methods.

Gearbox fault diagnosis based on RGT-MFFIN and multi-sensor fusion image generation

Abstract In gearbox fault diagnosis based on vibration and torque state data, traditional one-dimensional time-frequency domain analysis methods often suffer from insufficient feature expression and mining, and require complex noise reduction and filtering preprocessing. To address this issue, this paper proposes a fusion image generation method that integrates the advantages of recurrence plot (RP) and Gramian angular summation field (GASF) to generate recurrence Gramian transformed (RGT) images. This approach integrates both global and local fault information, making the fault characteristics more intuitive and easier to analyze. Given that multi-sensor collaboration can enhance feature representation, feature-level fusion increases the computational burden, and decision-level fusion is prone to losing inter-sensor correlation information, this paper adopts data-level fusion for image sample enhancement. In the diagnostic method, the challenge of traditional convolutional neural networks (CNNs) in extracting diverse geometric linear structures from fused images is addressed by introducing deformable convolutional blocks for initial feature extraction. Additionally, a multi-scale feature fusion interaction network (MFFIN) is constructed. This network incorporates a channel-space interactive attention mechanism on top of multi-scale feature extraction, assigning weights to features according to their importance while facilitating the interaction of feature information. Finally, validation is carried out using public datasets, and the experimental results show that the proposed method demonstrates significant advantages in classification accuracy and robustness under variable operating conditions and noise, thereby proving its effectiveness and practicality.

GT-Net: global transformer network for multiclass brain tumor classification using MR images.

Multiclass classification of brain tumors from magnetic resonance (MR) images is challenging due to high inter-class similarities. To this end, convolution neural networks (CNN) have been widely adopted in recent studies. However, conventional CNN architectures fail to capture the small lesion patterns of brain tumors. To tackle this issue, in this paper, we propose a global transformer network dubbed GT-Net for multiclass brain tumor classification. The GT-Net mainly comprises a global transformer module (GTM), which is introduced on the top of a backbone network. A generalized self-attention block (GSB) is proposed to capture the feature inter-dependencies not only across spatial dimension but also channel dimension, thereby facilitating the extraction of the detailed tumor lesion information while ignoring less important information. Further, multiple GSB heads are used in GTM to leverage global feature dependencies. We evaluate our GT-Net on a benchmark dataset by adopting several backbone networks, and the results demonstrate the effectiveness of GTM. Further, comparison with state-of-the-art methods validates the superiority of our model.

Complex Pseudo 3D Auto-Correlation Network for High-Quality Single-Angle Plane Wave Imaging

Abstract Deep learning has shown potential as an effective beamformer for improving the image quality of plane wave imaging (PWI). But most existing deep learning methods cannot directly handle the complex in-phase and quadrature (IQ) data. And noise in ultrasound signals would significantly damage the performance of regular convolution networks. To address these challenges, we proposed the Complex Pseudo 3D Auto-Correlation Network (CP3AN) which combined complex convolution and pseudo 3D auto-correlation blocks (P3AB) to directly map delayed IQ data from 0° plane wave into PWI pixels. The complex convolution could fully utilize the envelope and phase information of IQ data, while the P3AB used 1D convolution to extract noise in channel and space dimensions, allowing the network to prioritize valid signals with extremely low computational cost. We evaluated the performance of CP3AN through numerical simulations, phantom experiments, and in-vivo experiments, which showed comparable metrics with minimum variance (MV), including contrast (CR), contrast-to-noise ratio (CNR), generalized contrast-to-noise ratio (GCNR), lateral, and axial full-width half maximum (FWHM) at -9.60 dB, 1.12, 0.65, 319 um, and 344 um, respectively. The CP3AN achieved a low computational cost of 0.11 M floating-point operations (FLOPs), significantly lower than the MV or other compared deep learning-based methods. Our proposed method provided a promising solution for improving single-angle PWI imaging, particularly in situations where high frame rates are necessary.

LDMNet: Enhancing the Segmentation Capabilities of Unmanned Surface Vehicles in Complex Waterway Scenarios

Semantic segmentation-based Complex Waterway Scene Understanding has shown great promise in the environmental perception of Unmanned Surface Vehicles. Existing methods struggle with estimating the edges of obstacles under conditions of blurred water surfaces. To address this, we propose the Lightweight Dual-branch Mamba Network (LDMNet), which includes a CNN-based Deep Dual-branch Network for extracting image features and a Mamba-based fusion module for aggregating and integrating global information. Specifically, we improve the Deep Dual-branch Network structure by incorporating multiple Atrous branches for local fusion; we design a Convolution-based Recombine Attention Module, which serves as the gate activation condition for Mamba-2 to enhance feature interaction and global information fusion from both spatial and channel dimensions. Moreover, to tackle the directional sensitivity of image serialization and the impact of the State Space Model’s forgetting strategy on non-causal data modeling, we introduce a Hilbert curve scanning mechanism to achieve multi-scale feature serialization. By stacking feature sequences, we alleviate the local bias of Mamba-2 towards image sequence data. LDMNet integrates the Deep Dual-branch Network, Recombine Attention, and Mamba-2 blocks, effectively capturing the long-range dependencies and multi-scale global context information of Complex Waterway Scene images. The experimental results on four benchmarks show that the proposed LDMNet significantly improves obstacle edge segmentation performance and outperforms existing methods across various performance metrics.

Semantic Interaction Meta-Learning Based on Patch Matching Metric.

Metric-based meta-learning methods have demonstrated remarkable success in the domain of few-shot image classification. However, their performance is significantly contingent upon the choice of metric and the feature representation for the support classes. Current approaches, which predominantly rely on holistic image features, may inadvertently disregard critical details necessary for novel tasks, a phenomenon known as "supervision collapse". Moreover, relying solely on visual features to characterize support classes can prove to be insufficient, particularly in scenarios involving limited sample sizes. In this paper, we introduce an innovative framework named Patch Matching Metric-based Semantic Interaction Meta-Learning (PatSiML), designed to overcome these challenges. To counteract supervision collapse, we have developed a patch matching metric strategy based on the Transformer architecture to transform input images into a set of distinct patch embeddings. This approach dynamically creates task-specific embeddings, facilitated by a graph convolutional network, to formulate precise matching metrics between the support classes and the query image patches. To enhance the integration of semantic knowledge, we have also integrated a label-assisted channel semantic interaction strategy. This strategy merges word embeddings with patch-level visual features across the channel dimension, utilizing a sophisticated language model to combine semantic understanding with visual information. Our empirical findings across four diverse datasets reveal that the PatSiML method achieves a classification accuracy improvement of 0.65% to 21.15% over existing methodologies, underscoring its robustness and efficacy.

Improved MobileNet V3-Based Identification Method for Road Adhesion Coefficient.

To enable the timely adjustment of the control strategy of automobile active safety systems, enhance their capacity to adapt to complex working conditions, and improve driving safety, this paper introduces a new method for predicting road surface state information and recognizing road adhesion coefficients using an enhanced version of the MobileNet V3 model. On one hand, the Squeeze-and-Excitation (SE) is replaced by the Convolutional Block Attention Module (CBAM). It can enhance the extraction of features effectively by considering both spatial and channel dimensions. On the other hand, the cross-entropy loss function is replaced by the Bias Loss function. It can reduce the random prediction problem occurring in the optimization process to improve identification accuracy. Finally, the proposed method is evaluated in an experiment with a four-wheel-drive ROS robot platform. Results indicate that a classification precision of 95.53% is achieved, which is higher than existing road adhesion coefficient identification methods.

Multi-Task Environmental Perception Methods for Autonomous Driving.

In autonomous driving, environmental perception technology often encounters challenges such as false positives, missed detections, and low accuracy, particularly in detecting small objects and complex scenarios. Existing algorithms frequently suffer from issues like feature redundancy, insufficient contextual interaction, and inadequate information fusion, making it difficult to perform multi-task detection and segmentation efficiently. To address these challenges, this paper proposes an end-to-end multi-task environmental perception model named YOLO-Mg, designed to simultaneously perform traffic object detection, lane line detection, and drivable area segmentation. First, a multi-stage gated aggregation network (MogaNet) is employed during the feature extraction process to enhance contextual interaction by improving diversity in the channel dimension, thereby compensating for the limitations of feed-forward neural networks in contextual understanding. Second, to further improve the model's accuracy in detecting objects of various scales, a restructured weighted bidirectional feature pyramid network (BiFPN) is introduced, optimizing cross-level information fusion and enabling the model to handle object detection at different scales more accurately. Finally, the model is equipped with one detection head and two segmentation heads to achieve efficient multi-task environmental perception, ensuring the simultaneous execution of multiple tasks. The experimental results on the BDD100K dataset demonstrate that the model achieves a mean average precision (mAP50) of 81.4% in object detection, an Intersection over Union (IoU) of 28.9% in lane detection, and a mean Intersection over Union (mIoU) of 92.6% in drivable area segmentation. The tests conducted in real-world scenarios show that the model performs effectively, significantly enhancing environmental perception in autonomous driving and laying a solid foundation for safer and more reliable autonomous driving systems.

GINet:Graph interactive network with semantic-guided spatial refinement for salient object detection in optical remote sensing images

There are many challenging scenarios in the task of salient object detection in optical remote sensing images (RSIs), such as various scales and irregular shapes of salient objects, cluttered backgrounds, etc. Therefore, it is difficult to directly apply saliency models targeting natural scene images to optical RSIs. Besides, existing models often do not give sufficient exploration for the potential relationship of different salient objects or different parts of the salient object. In this paper, we propose a graph interaction network (i.e. GINet) with semantic-guided spatial refinement to conduct salient object detection in optical RSIs. The key advantages of GINet lie in two points. Firstly, the graph interactive reasoning (GIR) module conducts information exchange of different-level features via the graph interaction operation, and enhances features along spatial and channel dimensions via the graph reasoning operation. Secondly, we designed the global content-aware refinement (GCR) module, which incorporates the foreground and background feature-based local information and the semantic feature-based global information simultaneously. Experiments results on two public optical RSIs datasets clearly show the effectiveness and superiority of the proposed GINet when compared with the state-of-the-art models.