CNN-based Network Research Articles

Semi-supervised Learning of a Vision Transformer for the Task of Road Traffic Segmentation in an Unstructured Environment

In the last few years, traditionally used for natural language processing tasks, recurrent neural networks have been replaced mainly by transformers. Thanks to the novel attention mechanism, they also sequentially receive text input but provide much better results than LSTM, GRU-based, or similar networks. Self-attention negates the problem of fading memory by allowing efficient evaluation of dependencies between distant tokens and provides a better means for parallelization for modern processing units like GPU. Until recently, the use of transformers for computer vision (CV) tasks was minimal. The biggest obstacles that hindered the progress in this field were immense computational complexity, the fact that the image is a grid, not a sequence-like text, and the lack of strong inductive bias, in other words, the ability to have a good grasp of local correlations, unlike their CNN counterparts. The latest slowed down the vision transformer (ViT) usage rate in semantic segmentation (SS) even more. However, it was recently shown that with sufficient data, Transformers could outperform CNN-based networks in image classification and, with the proper ViT structure, even in SS. A promising direction for providing a ViT with required training data is using semi-supervised learning (SSL), which allows for extracting helpful information from unlabeled data using only a small amount of labeled data. This approach is beneficial when solving the problem of SS since manually creating masks for images is very time-consuming. This paper proposes the robust semi-supervised ViT learning method using minimal labeled data. The combination of a strong augmentation pipeline and a dual teacher paradigm allows good performance for SS of road traffic in the unstructured environment without the need for extensive hyperparameter search

Intelligent Fault Diagnosis Across Varying Working Conditions Using Triplex Transfer LSTM for Enhanced Generalization

Fault diagnosis plays a pivotal role in ensuring the reliability and efficiency of industrial machinery. While various machine/deep learning algorithms have been employed extensively for diagnosing faults in bearings and gears, the scarcity of data and the limited availability of labels have become a major bottleneck in developing data-driven diagnosis approaches, restricting the accuracy of deep networks. To overcome the limitations of insufficient labeled data and domain shift problems, an intelligent, data-driven approach based on the Triplex Transfer Long Short-Term Memory (TTLSTM) network is presented, which leverages transfer learning and fine-tuning strategies. Our proposed methodology uses empirical mode decomposition (EMD) to extract pertinent features from raw vibrational signals and utilizes Pearson correlation coefficients (PCC) for feature selection. L2 regularization transfer learning is utilized to mitigate the overfitting problem and to improve the model’s adaptability in diverse working conditions, especially in scenarios with limited labeled data. Compared with traditional transfer learning approaches, such as TCA, BDA, and JDA, which demonstrate accuracies in the range of 40–50%, our proposed model excels in identifying machinery faults with minimal labeled data by achieving 99.09% accuracy. Moreover, it performs significantly better than classical methods like SVM, RF, and CNN-based networks found in the literature, demonstrating the improved performance of our approach in fault diagnosis under varying working conditions and proving its applicability in real-world applications.

Comparative Bladder Cancer Tissues Prediction Using Vision Transformer.

Bladder cancer, often asymptomatic in the early stages, is a type of cancer where early detection is crucial. Herein, endoscopic images are meticulously evaluated by experts, and sometimes even by different disciplines, to identify tissue types. It is believed that the time spent by experts can be utilized for patient treatment with the creation of a computer-aided decision support system. For this purpose, in this study, it is evaluated that the performances of three models proposed using the bladder tissue dataset. The first model is a convolutional neural network (CNN)-based deep learning (DL) network, and the second is a model named hybrid cnn-machine learning (ML) or DL + ML, which involves classifying deep features obtained from a CNN-based network with ML. The last one, and the one that achieved the best performance metrics, is a vision transformer (ViT) architecture. Furthermore, a graphical user interface (GUI) is provided for an accessible decision support system. As a result, accuracy and F1 score values for DL, DL + ML, and ViT models are 0.9086-0.8971-0.9257 and 0.8884-0.8496-0.8931, respectively.

Development of a transformer-based deep learning algorithm for diabetic peripheral neuropathy classification using corneal confocal microscopy images.

Diabetic peripheral neuropathy (DPN) is common and can go unnoticed until it is firmly developed. This study aims to establish a transformer-based deep learning algorithm (DLA) to classify corneal confocal microscopy (CCM) images, identifying DPN in diabetic patients. Our classification model differs from traditional convolutional neural networks (CNNs) using a Swin transformer network with a hierarchical architecture backbone. Participants included those with (DPN+, n = 57) or without (DPN-, n = 37) DPN as determined by the updated Toronto consensus criteria. The CCM image dataset (consisting of 570 DPN+ and 370 DPN- images, with five images selected from each participant's left and right eyes) was randomly divided into training, validation, and test subsets at a 7:1:2 ratio, considering individual participants. The effectiveness of the algorithm was assessed using diagnostic accuracy measures, such as sensitivity, specificity, and accuracy, in conjunction with Grad-CAM visualization techniques to interpret the model's decisions. In the DPN + group (n = 12), the transformer model successfully predicted all participants, while in the DPN- group (n = 7), one participant was misclassified as DPN+, with an area under the curve (AUC) of 0.9405 (95% CI 0.8166, 1.0000). Among the DPN + images (n = 120), 117 were correctly classified, and among the DPN- images (n = 70), 49 were correctly classified, with an AUC of 0.8996 (95% CI 0.8502, 0.9491). For single-image predictions, the transformer model achieved a superior AUC relative to the ResNet50 model (0.8761, 95% CI 0.8155, 0.9366), the Inception_v3 model (0.8802, 95% CI 0.8231, 0.9374), and the DenseNet121 model (0.8965, 95% CI 0.8438, 0.9491). Transformer-based networks outperform CNN-based networks in rapid binary DPN classification. Transformer-based DLAs have clinical DPN screening potential.

Accelerated muscle mass estimation from CT images through transfer learning

BackgroundThe cost of labeling to collect training data sets using deep learning is especially high in medical applications compared to other fields. Furthermore, due to variances in images depending on the computed tomography (CT) devices, a deep learning based segmentation model trained with a certain device often does not work with images from a different device.MethodsIn this study, we propose an efficient learning strategy for deep learning models in medical image segmentation. We aim to overcome the difficulties of segmentation in CT images by training a VNet segmentation model which enables rapid labeling of organs in CT images with the model obtained by transfer learning using a small number of manually labeled images, called SEED images. We established a process for generating SEED images and conducting transfer learning a model. We evaluate the performance of various segmentation models such as vanilla UNet, UNETR, Swin-UNETR and VNet. Furthermore, assuming a scenario that a model is repeatedly trained with CT images collected from multiple devices, in which is catastrophic forgetting often occurs, we examine if the performance of our model degrades.ResultsWe show that transfer learning can train a model that does a good job of segmenting muscles with a small number of images. In addition, it was confirmed that VNet shows better performance when comparing the performance of existing semi-automated segmentation tools and other deep learning networks to muscle and liver segmentation tasks. Additionally, we confirmed that VNet is the most robust model to deal with catastrophic forgetting problems.ConclusionIn the 2D CT image segmentation task, we confirmed that the CNN-based network shows better performance than the existing semi-automatic segmentation tool or latest transformer-based networks.

Quantitative strategies of different loss functions aggregation for knowledge distillation

<p>Deep learning models have been successfully applied to many visual tasks. However, they tend to be increasingly cumbersome due to their high computational complexity and large storage requirements. How to compress convolutional neural network (CNN) models while still maintain their efficiency has received increasing attention from the community, and knowledge distillation (KD) is efficient way to do this. Existing KD methods have focused on the selection of good teachers from multiple teachers, or KD layers, which is cumbersome, expensive computationally, and requires large neural networks for individual models. Most of teacher and student modules are CNN-based networks. In addition, recent proposed KD methods have utilized cross entropy (CE) loss function at student network and KD network. This research focuses on the quantifiable evaluation of teacher-student model, in which knowledge is not only distilled from training models that have the same CNN architecture but also from different architectures. Furthermore, we propose combination of CE, balance cross entropy (BCE), and focal loss functions to not only soften the value of loss function in transferring knowledge from large teacher model to small student model but also increase classification performance. The proposed solution is evaluated on four benchmark static image datasets, and the experimental results show that our proposed solution outperforms the state-of-the-art (SOTA) methods from 2.67% to 9.84% at top 1 accuracy.</p>

CNN-based Network Intrusion Detection and Classification Model for Cyber-Attacks

A Convolution Neural Network (CNN)-based Network Intrusion Detection Model for Cyber-attacks is of great value in identifying and classifying attacks on any network. The Knowledge Discovery in Database Cup '99 dataset containing approximately 4,900,000 single connection vectors was divided into two phases; 75% of the total dataset was used during the learning process of the machine learning technique, while 25% was used on a fully trained model to validate and evaluate its performance. The model's performance indicated that it can detect and classify different classes of attacks with an accuracy of 98% with 20 epochs at a 0.001 learning rate using machine learning. The model loss for the training and validation was 7.48% and 7.98%, respectively, over 20 epochs, which implies that the model performed better on the training dataset. This study demonstrated that the convolutional Neural network-based Network Intrusion Detection and classification model shows high detection and low false negative rates. The CNN model offers a high detection rate and fidelity to unknown attacks, i.e., it can differentiate between already-seen attacks and new zero-day attacks. At the end of the experiment, the proposed approach is suitable in modeling the network IDS for detecting intrusion attacks on computer networks thereby enabling a secured environment for the proper functioning of the system

Parallel diffusion models promote high detail-fidelity photoacoustic microscopy in sparse sampling.

Reconstructing sparsely sampled data is fundamental for achieving high spatiotemporal resolution photoacoustic microscopy (PAM) of microvascular morphology in vivo. Convolutional networks (CNN) and generative adversarial networks (GAN) have been introduced to high-speed PAM, but due to the use of upsampling in CNN-based networks to restore details and the instability in GAN training, they struggle to learn the entangled microvascular network structure and vascular texture features, resulting in only achieving low detail-fidelity imaging of microvascular. The diffusion models is richly sampled and can generate high-quality images, which is very helpful for the complex vascular features in PAM. Here, we propose an approach named parallel diffusion models (PDM) with parallel learning of Noise task and Image task, where the Noise task optimizes through variational lower bounds to generate microvascular structures that are visually realistic, and the Image task improves the fidelity of the generated microvascular details through image-based loss. With only 1.56% of fully sampled pixels from photoacoustic human oral data, PDM achieves an LPIPS of 0.199. Additionally, using PDM in high-speed 16x PAM prevents breathing artifacts and image distortion issues caused by low-speed sampling, reduces the standard deviation of the Row-wise Self-Correlation Coefficient, and maintains high image quality. It achieves high confidence in reconstructing detailed information from sparsely sampled data and will promote the application of reconstructed sparsely sampled data in realizing high spatiotemporal resolution PAM.

Classification of osteoarthritic and healthy cartilage using deep learning with Raman spectra

Raman spectroscopy is a rapid method for analysing the molecular composition of biological material. However, noise contamination in the spectral data necessitates careful pre-processing prior to analysis. Here we propose an end-to-end Convolutional Neural Network to automatically learn an optimal combination of pre-processing strategies, for the classification of Raman spectra of superficial and deep layers of cartilage harvested from 45 Osteoarthritis and 19 Osteoporosis (Healthy controls) patients. Using 6-fold cross-validation, the Multi-Convolutional Neural Network achieves comparable or improved classification accuracy against the best-performing Convolutional Neural Network applied to either the raw or pre-processed spectra. We utilised Integrated Gradients to identify the contributing features (Raman signatures) in the network decision process, showing they are biologically relevant. Using these features, we compared Artificial Neural Networks, Decision Trees and Support Vector Machines for the feature selection task. Results show that training on fewer than 3 and 300 features, respectively, for the disease classification and layer assignment task provide performance comparable to the best-performing CNN-based network applied to the full dataset. Our approach, incorporating multi-channel input and Integrated Gradients, can potentially facilitate the clinical translation of Raman spectroscopy-based diagnosis without the need for laborious manual pre-processing and feature selection.

Points of interest linear attention network for real-time non-rigid liver volume to surface registration.

In laparoscopic liver surgery, accurately predicting the displacement of key intrahepatic anatomical structures is crucial for informing the doctor's intraoperative decision-making. However, due to the constrained surgical perspective, only a partial surface of the liver is typically visible. Consequently, the utilization of non-rigid volume to surface registration methods becomes essential. But traditional registration methods lack the necessary accuracy and cannot meet real-timerequirements. To achieve high-precision liver registration with only partial surface information and estimate the displacement of internal liver tissues inreal-time. We propose a novel neural network architecture tailored for real-time non-rigid liver volume to surface registration. The network utilizes a voxel-based method, integrating sparse convolution with the newly proposed points of interest (POI) linear attention module. POI linear attention module specifically calculates attention on the previously extracted POI. Additionally, we identified the most suitable normalization methodRMSINorm. We evaluated our proposed network and other networks on a dataset generated from real liver models and two real datasets. Our method achieves an average error of 4.23mm and a mean frame rate of 65.4fps in the generation dataset. It also achieves an average error of 8.29mm in the human breathing motiondataset. Our network outperforms CNN-based networks and other attention networks in terms of accuracy and inferencespeed.

TriConvUNeXt: A Pure CNN-Based Lightweight Symmetrical Network for Biomedical Image Segmentation.

Biomedical image segmentation is essential in clinical practices, offering critical insights for accurate diagnosis and strategic treatment approaches. Nowadays, self-attention-based networks have achieved competitive performance in both natural language processing and computer vision, but the computational cost has reduced their popularity in practical applications. The recent study of Convolutional Neural Network (CNN) explores linear functions within modified CNN layer demonstrating pure CNN-based networks can still achieve competitive results against Vision Transformer (ViT) in biomedical image segmentation, with fewer parameters. The modified CNN, i.e., Depthwise CNN, however, leaves the features cleaved off in the channel dimension and prevents the extraction of features in the perspective of channel interaction. To effectively further explore the feature learning power of modified CNN with biomedical image segmentation, we design a lightweight multi-convolutional multi-scale convolutional network block (MSConvNeXt) for U-shape symmetrical network. Specifically, a network block consisting of a depthwise CNN, a deformable CNN, and a dilated CNN is proposed to capture semantic feature information effectively while with low computational cost. Furthermore, channel shuffling operation is proposed to dynamically promote an efficient feature fusion among the feature maps. The network block hereby is properly deployed within U-shape symmetrical encoder-decoder style network, named TriConvUNeXt. The proposed network is validated on a public benchmark dataset with a comprehensive evaluation in both computational cost and segmentation performance against 13 baseline methods. Specifically, TriConvUNeXt achieves 1% higher than UNet and TransUNet in Dice-Coefficient while 81% and 97% lower in computational cost. The implementation of TriConvUNeXt is made publicly accessible via https://github.com/ziyangwang007/TriConvUNeXt .

Crack detection and dimensional assessment using smartphone sensors and deep learning

This paper addresses the crucial need for effective crack detection and dimensional assessment in civil infrastructure materials to ensure safety and functionality. It proposes a cost-effective crack detection and dimensional assessment solution by applying state-of-the-art deep learning on smartphone sensor imagery and positioning data. The proposed methodology integrates three-dimensional (3D) data from LiDAR sensors with Mask R-convolutional neural network (CNN) and YOLOv8 object detection networks, for automated crack detection in concrete structures, allowing for accurate measurement of crack dimensions, including length, width, and area. The study finds that YOLOv8 produces superior precision and recall results in crack detection compared to Mask R-CNN. Furthermore, the calculated crack-straight-length closely aligns with the ground-truth straight-length, with an average error of 1.5%. These research contributions include developing a multi-modal solution combining LiDAR observations with image masks for precise 3D crack measurements, establishing a dimensional assessment pipeline to convert segmented cracks into measurements, and comparing state-of-the-art CNN-based networks for crack detection in real-life images.

IWNeXt: an image-wavelet domain ConvNeXt-based network for self-supervised multi-contrast MRI reconstruction

Objective. Multi-contrast magnetic resonance imaging (MC MRI) can obtain more comprehensive anatomical information of the same scanning object but requires a longer acquisition time than single-contrast MRI. To accelerate MC MRI speed, recent studies only collect partial k-space data of one modality (target contrast) to reconstruct the remaining non-sampled measurements using a deep learning-based model with the assistance of another fully sampled modality (reference contrast). However, MC MRI reconstruction mainly performs the image domain reconstruction with conventional CNN-based structures by full supervision. It ignores the prior information from reference contrast images in other sparse domains and requires fully sampled target contrast data. In addition, because of the limited receptive field, conventional CNN-based networks are difficult to build a high-quality non-local dependency. Approach. In the paper, we propose an Image-Wavelet domain ConvNeXt-based network (IWNeXt) for self-supervised MC MRI reconstruction. Firstly, INeXt and WNeXt based on ConvNeXt reconstruct undersampled target contrast data in the image domain and refine the initial reconstructed result in the wavelet domain respectively. To generate more tissue details in the refinement stage, reference contrast wavelet sub-bands are used as additional supplementary information for wavelet domain reconstruction. Then we design a novel attention ConvNeXt block for feature extraction, which can capture the non-local information of the MC image. Finally, the cross-domain consistency loss is designed for self-supervised learning. Especially, the frequency domain consistency loss deduces the non-sampled data, while the image and wavelet domain consistency loss retain more high-frequency information in the final reconstruction. Main results. Numerous experiments are conducted on the HCP dataset and the M4Raw dataset with different sampling trajectories. Compared with DuDoRNet, our model improves by 1.651 dB in the peak signal-to-noise ratio. Significance. IWNeXt is a potential cross-domain method that can enhance the accuracy of MC MRI reconstruction and reduce reliance on fully sampled target contrast images.

ViComp: Video Compensation for Projector-Camera Systems.

Projector video compensation aims to cancel the geometric and photometric distortions caused by non-ideal projection surfaces and environments when projecting videos. Most existing projector compensation methods start by projecting and capturing a set of sampling images, followed by an offline compensation model training step. Thus, abundant user effort is required before the users can watch the video. Moreover, the sampling images have little prior knowledge of the video content and may lead to suboptimal results. To address these issues, this paper builds a video compensation system that can online adapt the compensation parameters. Our approach consists of five threads and can perform compensation, projection, capturing, and short-term and long-term model updates in parallel. Due to the parallel mechanism, rather than projecting and capturing hundreds of sampling images and training the model offline, we can directly use the projected and captured video frames for model updates on the fly. To quickly apply to the new environment, we introduce a deep learning-based compensation model that integrates a fixed transformer-based method and a novel CNN-based network. Moreover, for fast convergence and to reduce error accumulation during fine-tuning, we present a strategy that cooperates with short-term and long-term memory model updates. Experiments show that it significantly outperforms state-of-the-art baselines.

CREVNet: A Transformer and CNN-Based Network for Accurate Segmentation of Ice Shelf Crevasses

CREVNet: A Transformer and CNN-Based Network for Accurate Segmentation of Ice Shelf Crevasses

A CNN-based network with attention mechanism for autonomous crack identification on building facade

ABSTRACT This paper presents a rapid and precise deep learning-based approach for measuring cracks in concrete structures. The proposed methodology involves data acquisition, pre-processing, model construction and post-processing. The Deeplabv3+ model is used for crack detection, and the use of Coordinate Attention is introduced to enhance its performance. And the implications of attention locations in the model are discussed. Insertion position after atrous spatial pyramid pooling (ASPP) operation is the most effective and accurate. The crack widths are obtained through post-processing, and the actual width is determined using the width conversion equation. The proposed method achieves an effective crack detection MIoU of 85.95 and a calculated width error of 6.3%, with a reduction of 2.5% compared to traditional models. Numerical experiments and real-world building experiments have demonstrated the feasibility and effectiveness of the proposed method. Overall, the proposed method presents a fast and dependable crack detection technique that has great potential for practical application.

Deep learning-based size prediction for optical trapped nanoparticles and extracellular vesicles from limited bandwidth camera detection.

Due to its ability to record position, intensity, and intensity distribution information, camera-based monitoring of nanoparticles in optical traps can enable multi-parametric morpho-optical characterization at the single-particle level. However, blurring due to the relatively long (10s of microsecond) integration times and aliasing from the resulting limited temporal bandwidth affect the detected particle position when considering nanoparticles in traps with strong stiffness, leading to inaccurate size predictions. Here, we propose a ResNet-based method for accurate size characterization of trapped nanoparticles, which is trained by considering only simulated time series data of nanoparticles' constrained Brownian motion. Experiments prove the method outperforms state-of-art sizing algorithms such as adjusted Lorentzian fitting or CNN-based networks on both standard nanoparticles and extracellular vesicles (EVs), as well as maintains good accuracy even when measurement times are relatively short (<1s per particle). On samples of clinical EVs, our network demonstrates a well-generalized ability to accurately determine the EV size distribution, as confirmed by comparison with gold-standard nanoparticle tracking analysis (NTA). Furthermore, by combining the sizing network with still frame images from high-speed video, the camera-based optical tweezers have the unique capacity to quantify both the size and refractive index of bio-nanoparticles at the single-particle level. These experiments prove the proposed sizing network as an ideal path for predicting the morphological heterogeneity of bio-nanoparticles in optical potential trapping-related measurements.

Empowering Vision Transformer by Network Hyper-Parameter Selection for Whole Pelvis Prostate Planning Target Volume Auto-Segmentation

U-Net, based on a deep convolutional network (CNN), has been clinically used to auto-segment normal organs, while still being limited to the planning target volume (PTV) segmentation. This work aims to address the problems in two aspects: 1) apply one of the newest network architectures such as vision transformers other than the CNN-based networks, and 2) find an appropriate combination of network hyper-parameters with reference to recently proposed nnU-Net (“no-new-Net”). VT U-Net was adopted for auto-segmenting the whole pelvis prostate PTV as it consisted of fully transformer architecture. The upgraded version (v.2) applied the nnU-Net-like hyper-parameter optimizations, which did not fully cover the transformer-oriented hyper-parameters. Thus, we tried to find a suitable combination of two key hyper-parameters (patch size and embedded dimension) for 140 CT scans throughout 4-fold cross validation. The VT U-Net v.2 with hyper-parameter tuning yielded the highest dice similarity coefficient (DSC) of 82.5 and the lowest 95% Haussdorff distance (HD95) of 3.5 on average among the seven recently proposed deep learning networks. Importantly, the nnU-Net with hyper-parameter optimization achieved competitive performance, although this was based on the convolution layers. The network hyper-parameter tuning was demonstrated to be necessary even for the newly developed architecture of vision transformers.

Deep learning models for predicting the position of the head on an X-ray image for Cephalometric analysis

Cephalometric analysis is used to identify problems in the development of the skull, evaluate their treatment, and plan for possible surgical interventions. The paper aims to develop a Convolutional Neural Network that will analyze the head position on an X-ray image. It takes place in such a way that it recognizes whether the image is suitable and, if not, suggests a change in the position of the head for correction. This paper addresses the exact rotation of the head with a change in the range of a few degrees of rotation. The objective is to predict the correct head position to take an X-ray image for further Cephalometric analysis. The changes in the degree of rotations were categorized into 5 classes. Deep learning models predict the correct head position for Cephalometric analysis. An X-ray image dataset on the head is generated using CT scan images. The generated images are categorized into 5 classes based on a few degrees of rotations. A set of four deep-learning models were then used to generate the generated X-Ray images for analysis. This research work makes use of four CNN-based networks. These networks are trained on a dataset to predict the accurate head position on generated X-Ray images for analysis. Two networks of VGG-Net, one is the U-Net and the last is of the ResNet type. The experimental analysis ascertains that VGG-4 outperformed the VGG-3, U-Net, and ResNet in estimating the head position to take an X-ray on a test dataset with a measured accuracy of 98%. It is due to the incorrectly classified images are classified that are directly adjacent to the correct ones at intervals and the misclassification rate is significantly reduced.

Swin Transformer-Based Edge Guidance Network for RGB-D Salient Object Detection.

Salient object detection (SOD), which is used to identify the most distinctive object in a given scene, plays an important role in computer vision tasks. Most existing RGB-D SOD methods employ a CNN-based network as the backbone to extract features from RGB and depth images; however, the inherent locality of a CNN-based network limits the performance of CNN-based methods. To tackle this issue, we propose a novel Swin Transformer-based edge guidance network (SwinEGNet) for RGB-D SOD in which the Swin Transformer is employed as a powerful feature extractor to capture the global context. An edge-guided cross-modal interaction module is proposed to effectively enhance and fuse features. In particular, we employed the Swin Transformer as the backbone to extract features from RGB images and depth maps. Then, we introduced the edge extraction module (EEM) to extract edge features and the depth enhancement module (DEM) to enhance depth features. Additionally, a cross-modal interaction module (CIM) was used to integrate cross-modal features from global and local contexts. Finally, we employed a cascaded decoder to refine the prediction map in a coarse-to-fine manner. Extensive experiments demonstrated that our SwinEGNet achieved the best performance on the LFSD, NLPR, DES, and NJU2K datasets and achieved comparable performance on the STEREO dataset compared to 14 state-of-the-art methods. Our model achieved better performance compared to SwinNet, with 88.4% parameters and 77.2% FLOPs. Our code will be publicly available.