SE Block Research Articles

High-noise solar panel defect identification method based on the improved EfficientNet-V2

As a crucial element in photovoltaic power generation systems, the condition of solar panels significantly impacts the efficiency of power generation. The ability to accurately and promptly detect defects in solar panels is essential for enhancing system performance. This study introduces a novel model for identifying defects in photovoltaic modules, leveraging an enhanced version of EfficientNet-V2. This model aims to address challenges in identifying defects in infrared images of solar panels under conditions of high-noise and low-model efficiency. To address the challenges of high image noise and blur, this article initially presents a methodology that combines the Db4 wavelet transform with a blind deconvolution algorithm for comprehensive preprocessing of the original image. Furthermore, this study optimizes the model's feature representation capabilities by implementing key transformations within the EfficientNet-V2 network framework. Notably, we replaced the traditional SE block with the more efficient channel attention (ECA) mechanism module. Due to its lightweight structure and effective performance, ECA substantially improves the model's capacity to extract complex and abstract image features, while also accelerating the training process's convergence speed and enhancing overall computational efficiency. At the classifier level, this paper innovatively integrates the XGBoost ensemble learning algorithm into the model, substituting the conventional softmax classifier used in traditional convolutional neural network (CNN). With its superior generalization capabilities, robust nonlinear modeling skills, and efficient computational characteristics, XGBoost can more accurately detect minute defects in solar panels based on the deep features produced by EfficientNet-V2, thereby significantly improving the accuracy and robustness of defect detection. Simulation results demonstrate that the proposed model structure outperforms traditional CNN models in terms of accuracy and stability, underscoring the efficacy of the enhanced EfficientNet-V2 model in detecting solar panel defects under high-noise conditions.

Bengali-Sign: A Machine Learning-Based Bengali Sign Language Interpretation for Deaf and Non-Verbal People.

Sign language is undoubtedly a common way of communication among deaf and non-verbal people. But it is not common among hearing people to use sign language to express feelings or share information in everyday life. Therefore, a significant communication gap exists between deaf and hearing individuals, despite both groups experiencing similar emotions and sentiments. In this paper, we developed a convolutional neural network-squeeze excitation network to predict the sign language signs and developed a smartphone application to provide access to the ML model to use it. The SE block provides attention to the channel of the image, thus improving the performance of the model. On the other hand, the smartphone application brings the ML model close to people so that everyone can benefit from it. In addition, we used the Shapley additive explanation to interpret the black box nature of the ML model and understand the models working from within. Using our ML model, we achieved an accuracy of 99.86% on the KU-BdSL dataset. The SHAP analysis shows that the model primarily relies on hand-related visual cues to predict sign language signs, aligning with human communication patterns.

Retinal Blood Vessels Segmentation With Improved SE‐UNet Model

ABSTRACTAccurate segmentation of retinal vessels is crucial for the early diagnosis and treatment of eye diseases, for example, diabetic retinopathy, glaucoma, and macular degeneration. Due to the intricate structure of retinal vessels, it is essential to extract their features with precision for the semantic segmentation of medical images. In this study, an improved deep learning neural network was developed with a focus on feature extraction based on the U‐Net structure. The enhanced U‐Net combines the architecture of convolutional neural networks (CNNs) with SE blocks (squeeze‐and‐excitation blocks) to adaptively extract image features after each U‐Net encoder's convolution. This approach aids in suppressing nonvascular regions and highlighting features for specific segmentation tasks. The proposed method was trained and tested on the DRIVECHASE_DB1 and STARE datasets. As a result, the proposed model had an algorithmic accuracy, sensitivity, specificity, Dice coefficient (Dc), and Matthews correlation coefficient (MCC) of 95.62/0.9853/0.9652, 0.7751/0.7976/0.7773, 0.9832/0.8567/0.9865, 82.53/87.23/83.42, and 0.7823/0.7987/0.8345, respectively, outperforming previous methods, including UNet++, attention U‐Net, and ResUNet. The experimental results demonstrated that the proposed method improved the retinal vessel segmentation performance.

Internet of Things intrusion detection: Research and practice of NSENet and LSTM fusion models

To address the problems of complex environment, limited device computational resources and limited memory resources in the existing IoT, SELSTM, an intrusion detection system composed of NSENet and LSTM fusion based on SENet, is investigated. The NSENet part of the SELSTM system is based on the squeeze-and-excitation network (SENet). The lightweight computational modules NonLocal, SKConv and inverted residuals are fused into SE blocks, and self-attention of Nonlocal is used to improve the local receptive field of feature extraction. The channel attention and spatial attention of each part of the data are strengthened by the use of SKConv To enhance the adaptive convolution ability of the model and ensure the completeness of the information, the properties of the inverted residual structure are used to ensure that the gradient of the model decreases steadily without gradient explosion or disappearance. For the problem of data imbalance, the dataset is randomly resampled using the weight resampling technique to improve the balance of the dataset to ensure that the final detection effect of the model is more effective and generalized, while the data flow is divided into two parts for processing, and the model parameters are optimized using the model gradient optimizer consisting of the optimizer Lion and the optimization function Lookahead. The model extracts the spatial and temporal features of the data through multidimensional extraction to ensure the completeness of the data feature information in multiple dimensions, thus obtaining better detection results. The results of the experiments comparing the SELSTM model with other models on the intrusion dataset show that the intrusion detection model has a higher detection precision and accuracy than the traditional deep learning intrusion detection model, which indicates that the SELSTM has better detection performance properties and better practicality and effectiveness on IoT devices.

Bearing fault diagnosis based on a multiple-constraint modal-invariant graph convolutional fusion network

Multisensor data fusion method can improve the accuracy of bearing fault diagnosis, in order to address the problems of single-sensor data types and the insufficient exploration of redundancy and complementarity between different modal data in most existing multisensor data fusion methods for bearing fault diagnosis, a bearing fault diagnosis method based on a Multiple-Constraint Modal-Invariant Graph Convolutional Fusion Network (MCMI-GCFN) is proposed in this paper. Firstly, a Convolutional Autoencoder (CAE) and Squeeze-and-Excitation Block (SE block) are used to extract features of raw current and vibration signals. Secondly, the model introduces source domain classifiers and domain discriminators to capture modal invariance between different modal data based on domain adversarial training, making use of the redundancy and complementarity between multimodal data. Then, the spatial aggregation property of Graph Convolutional Neural Networks (GCN) is utilized to capture the dependency relationship between current and vibration modes with similar time step features for accurately fusing contextual semantic information. Finally, the validation is conducted on the public bearing damage current and vibration dataset from Paderborn University. The experimental results showed that the delivered fusion method achieved a bearing fault diagnosis accuracy of 99.6 %, which was about 9 %–11.4 % better than that with nonfusion methods.

Aeroengine Remaining Life Prediction Using Feature Selection and Improved SE Blocks

Aeroengines use numerous sensors to detect equipment health and ensure proper operation. Currently, filtering useful sensor data and removing useless data is challenging in predicting the remaining useful life (RUL) of an aeroengine using deep learning. To reduce computational costs and improve prediction performance, we use random forest to evaluate the feature importance of sensor data. Based on the size of the feature corresponding to the Gini index, we select the appropriate sensor. This helps us to determine which sensor to use and ensures that the computational resources are not wasted on unnecessary sensors. Considering that the RUL of equipment changes in a progressively more complex manner as the equipment is used over time, we propose an improved squeeze and excitation block (SSE) and combine it with a convolutional neural network (CNN). By enhancing the feature selection ability of CNN through segmented squeeze and excitation block, the model can focus on important information within features to effectively improve prediction performance. We compared our experiments with other RUL experiments on the CMAPSS aeroengine dataset and then conducted ablation experiments to verify the critical role of the methods we used.

Enhancing mask detection performance based on YOLOv5 model optimization and attention mechanisms

Due to the COVID-19 pandemic, there has been a significant increase in the usage of masks, leading to more complex scenarios for mask detection techniques. This paper focuses on optimizing the performance of mask detection using the You Only Look Once (YOLO) v5 model. In this study, the yolov5 target detection model was employed for training the mask dataset. Diverse model improvement techniques were explored to enhance the model's capability to capture crucial features and differentiate masks from the background in complex scenarios. Finally, the modified model was compared with the earlier original target detection model to identify the most considerable performance gain. The CSPDarknet design with the TensorFlow framework is utilized in this study, and the Attention Mechanism module is implemented through the Keras library. The objective is to optimize the three feature layers between the backbone network and the neck by integrating multiple attention mechanisms. This will enable the model to more quickly and accurately capture important features when dealing with complex scenarios by adjusting the feature map weights. Additionally, in the feature pyramid network, shallow feature maps are fused with deeper feature maps in a certain order to determine the most efficient feature fusion method. Finally, this study identified the optimal combination of attention mechanism and feature fusion through ablation experiments. The results of the experiment demonstrate that the combination of SE block and shallow feature fusion (SE + FF2 model) can greatly enhance category confidence, leading to an improved model performance.

PVCsNet : A Specialized Artificial Intelligence-Based Model to Classify Premature Ventricular Contractions from ECG Images.

Premature ventricular complexes (PVCs) are irregularities in heart rhythm where the ventricles contract earlier than expected, disrupting the normal cardiac cycle. Identifying the origin of PVCs before surgery is crucial as it can reduce operation duration, lower radiation exposure, and potentially enhance ablation success rates. Current detection methods face limitations in accuracy and data processing, often requiring large datasets and complex interpretations. This study presents PVCsNet, a deep-learning network specifically designed for classifying premature ventricular complexes (PVCs) in ECG images. It incorporates residual structures and attention mechanisms to enhance classification performance. PVCsNet consists of four 3×3 convolutional layers as feature extractors, followed by residual connections and attention blocks. This design enables the network to map image features to class probability distributions, enhancing performance even with limited data. Our experimental results demonstrate that using the SE Block with MaxPool and a ratio of 4, PVCsNet achieves an overall accuracy of 94.49%, with high precision in critical categories and a moderate parameter size. We successfully categorize the data into six distinct classes based on their origin locations in the heart: right ventricular outflow tract (RVOT), left ventricular outflow tract (LVOT), papillary muscle (PM), valvular annulus (VA), summit, and His-Purkinje system (HPS). Among these, RVOT is the most common and crucial origin of PVCs. PM and HPS are also significant origins due to their clinical implications. This study demonstrates the potential of PVCsNet in clinical diagnostics, providing promising results in classifying ECG images and contributing to future medical research and diagnosis.

Time-frequency enhanced characterization method based on asymmetric image reconstruction autoencoder

The vibration signals of mechanical equipment are subject to the influence of complex and variable working conditions, often exhibiting non-smooth and non-linear characteristics. The conventional time-frequency (TF) analysis (TFA) method, which relies on energy concentration, is susceptible to noise and impact, making it challenging to accurately extract fault characteristics. To overcome these limitations, this paper proposes an innovative approach. In this paper, we introduce an asymmetric image reconstruction autoencoder model, which is based on two well-known TFA methods, namely, short-time Fourier transform (STFT) and synchroextracting transform (SET), effectively reducing noise and improving the TF energy concentration process through learning the mapping relationship between STFT and SET. To address the clarity issue in the reconstructed TF images, the paper incorporates a channel attention mechanism known as SE Block into the encoding-decoding structure. Additionally, a skip connection structure is introduced to aid in restoring the structural details of the reconstructed TF images. Moreover, an improved weighted joint loss function is proposed to adaptively enhance various types of TF distribution features. This enhancement ensures that different characteristics of TF distribution are adequately addressed during the reconstruction process. The proposed method is put to the test using both simulated signals and experimental signals from gearbox rolling bearing faults. The results demonstrate that compared to traditional TFA and post-processing methods, the proposed model exhibits superior capabilities in enhancing the TF characterization of multi-source time-varying signals. Furthermore, it demonstrates remarkable robustness to noise and can accurately extract instantaneous frequency. These findings point to the promising potential of this method for mechanical fault identification and diagnosis applications.

Development of continuous cuffless blood pressure prediction platform using enhanced 1-D SENet–LSTM

In this study, we propose a blood pressure estimation that employs a combined 1-D squeeze and excitation network (SENet-LSTM) architecture. Combining photoplethysmography (PPG) and electrocardiogram (ECG) signals can provide more helpful feature information. The original signals were processed by removing the noise and artifacts. By introducing SE block, the ability to learn features is enhanced, and an end-to-end approach is used to realize the automatic learning process. Additionally, the algorithm has the ability to recover arterial blood pressure (ABP) waveforms from blood pressure readings. The results of this study met the Grade A standard for diastolic blood pressure (DBP), systolic blood pressure (SBP), and mean arterial pressure (MAP) set by the British Hypertension Society (BHS) and were also in accordance with the Association for the Advancement of Medical Instrumentation (AAMI) standard. According to the DBP and SBP ranges, blood pressure is divided into the following categories: normal blood pressure, prehypertension, and hypertension. For the classification of Hypertension, Normotension, and Prehypertension using SBP, we achieved accuracy of 94% and F1 scores of 0.92, 0.94, and 0.85. For the results obtained using DBP classification the overall accuracy was 91%, with F1 scores of 0.85, 0.98, and 0.78. When compared to other studies, the classifier results generated by SBP and DBP estimates based on the 1-D SENet-LSTM method improved accuracy by 2% and 11%, respectively.

PLU-Net: Extraction of multiscale feature fusion.

In recent years, deep learning algorithms have achieved remarkable results in medical image segmentation. These networks with an enormous number of parameters often encounter challenges in handling image boundaries and details, which may result in suboptimal segmentation results. To solve the problem, we develop atrous spatial pyramid pooling (ASPP) and combine it with the squeeze-and-excitation block (SE block), as well as present the PS module, which employs a broader and multiscale receptive field at the network's bottom to obtain more detailed semantic information. We also propose the local guided block (LG block) and also its combination with the SE block to form the LS block, which can obtain more abundant local features in the feature map, so that more edge information can be retained in each down sampling process, thereby improving the performance of boundary segmentation. We propose PLU-Net and integrate our PS module and LS block into U-Net. We put our PLU-Net to the test on three benchmark datasets, and the results show that by fewer parameters and FLOPs, it outperforms on medical semantic segmentationtasks.

Multi-scale SE-residual network with transformer encoder for myocardial infarction classification

In recent years, there has been an increase in the number of deaths caused by heart disease. The design of computer-aided systems for the diagnosis of heart diseases, especially myocardial infarction (MI), has become a hot topic. The diagnosis of MI is a classification task. Electrocardiogram (ECG) is one of the most important tools for diagnosing heart disease. ECG signals are classified into corresponding MI categories based on their extracted features. Thus, the diagnosis of MI depends on the feature extraction of the ECG signal. Because signal processing methods, such as wavelet transform, may not effectively extract hidden features, researchers nowadays turn to using convolutional neural networks to solve the problem of hidden feature extraction. However, this approach ignores the role of temporal information. we proposed a new network called Multi-scale SE-Residual Network with Transformer encoder (MRTNet) to extract hidden features and temporal information features better. The data processed are localized to heartbeat units based on the way cardiologists diagnose myocardial infarction. Inspired by multi-scale learning, the sampling module was designed to acquire heartbeat units at different scales. We provided a processing model for multi-scale data, called the feature extraction module. The residual network with SE block and Customized Pooling Component (CPC) was used to extract global and local features of the heartbeat units. A Transformer encoder was used to extract the common features of the previous module’s output. The common features at different scales were fused to provide the basis for the classification task. The experiment was conducted on the public PTB-XL dataset. MRTNet achieved a higher accuracy and F1 score than other models by approximately 2%. Furthermore, MRTNet demonstrated a superior AUC of approximately 0.77. It is a superior classifier when compared to other models. The results demonstrate the effectiveness of the designed multi-scale architecture. And it provides a way into exploring further potential information from ECG signals.

A novel approach to attention mechanism using kernel functions: Kerformer.

Artificial Intelligence (AI) is driving advancements across various fields by simulating and enhancing human intelligence. In Natural Language Processing (NLP), transformer models like the Kerformer, a linear transformer based on a kernel approach, have garnered success. However, traditional attention mechanisms in these models have quadratic calculation costs linked to input sequence lengths, hampering efficiency in tasks with extended orders. To tackle this, Kerformer introduces a nonlinear reweighting mechanism, transforming maximum attention into feature-based dot product attention. By exploiting the non-negativity and non-linear weighting traits of softmax computation, separate non-negativity operations for Query(Q) and Key(K) computations are performed. The inclusion of the SE Block further enhances model performance. Kerformer significantly reduces attention matrix time complexity from O(N2) to O(N), with N representing sequence length. This transformation results in remarkable efficiency and scalability gains, especially for prolonged tasks. Experimental results demonstrate Kerformer's superiority in terms of time and memory consumption, yielding higher average accuracy (83.39%) in NLP and vision tasks. In tasks with long sequences, Kerformer achieves an average accuracy of 58.94% and exhibits superior efficiency and convergence speed in visual tasks. This model thus offers a promising solution to the limitations posed by conventional attention mechanisms in handling lengthy tasks.

Computer aided detection of leaf disease in agriculture using convolution neural network based squeeze and excitation network

ABSTRACT The support rendered by artificial intelligence in plant disease diagnosis and with drastic progression in the agricultural technology, it is necessary to do pertinent research for the cause of long-term agricultural development. Numerous diseases like early and late blight have a significant influence on the quality and quantity of potatoes. Manual interpretation turns out to be a time-consuming process in sorting out leaf diseases. In order to classify various diseases like fungal, viral and bacterial infections in the potato leaf, an enhanced Convolution Neural Network based on VGG16 is used for potato leaf disease classification. Improved Median filter is also used which eradicates the noise to a greater extent. The convolution layers of VGG16 along with the Inception and the SE block are used in this research for classification. The global average pooling layer is used to reduce model training parameters, layer and Squeeze and Excitation Network attention mechanism is used to improve the model’s ability to extract features. The approximate calculations can be done by using soft computing. Compared with other traditional convolutional neural networks, the proposed model achieved the highest classification accuracy of 99.3%

Micro-architecture design exploration template for AutoML case study on SqueezeSEMAuto

Convolutional Neural Network (CNN) models have been commonly used primarily in image recognition tasks in the deep learning area. Finding the right architecture needs a lot of hand-tune experiments which are time-consuming. In this paper, we exploit an AutoML framework that adds to the exploration of the micro-architecture block and the multi-input option. The proposed adaption has been applied to SqueezeNet with SE blocks combined with the residual block combinations. The experiments assume three search strategies: Random, Hyperband, and Bayesian algorithms. Such combinations can lead to solutions with superior accuracy while the model size can be monitored. We demonstrate the application of the approach against benchmarks: CIFAR-10 and Tsinghua Facial Expression datasets. The searches allow the designer to find the architectures with better accuracy than the traditional architectures without hand-tune efforts. For example, CIFAR-10, leads to the SqueezeNet architecture using only 4 fire modules with 59% accuracy. When exploring SE block insertion, the model with good insertion points can lead to an accuracy of 78% while the traditional SqueezeNet can achieve an accuracy of around 50%. For other tasks, such as facial expression recognition, the proposed approach can lead up to an accuracy of 71% with the proper insertion of SE blocks, the appropriate number of fire modules, and adequate input merging, while the traditional model can achieve the accuracy under 20%.

Information extraction from Visually Rich Documents using graph convolutional network

Visually rich documents, such as forms, invoices, receipts, and ID cards, are ubiquitous in daily business and life. Various methods have been used to convey such diverse information, including text, layout, font size, or text position. Combining these elements in information extraction can improve the result performance. However, previous works have not effectively utilized the cooperation between these rich information sources. Text detection and recognition have been performed without semantic supervision (e.g., entity name annotation), and text information extraction has been performed using only serialized plain text, ignoring rich visual information. This paper presents a method for extracting information from such documents, which integrates textual, non-spatial, and spatial visual features. The method consists of two main steps and uses three deep neural networks. The first step, Text Reading, employs two CNN models (Lightweight DB and C-PREN) for OCR tasks, based on the state-of-the-art models DB and PREN, with two improvements. These improvements include reducing noise by removing the SE block of DB and integrating both context and position features in PREN. The second step, Text Information Extraction, uses a graph convolutional network, RGCN, for name entity recognition. Experiments on self-collected and two public datasets have demonstrated that our method improves the performance of the original models and outperforms other state-of-the-art methods.

3D network with channel excitation and knowledge distillation for action recognition.

Modern action recognition techniques frequently employ two networks: the spatial stream, which accepts input from RGB frames, and the temporal stream, which accepts input from optical flow. Recent researches use 3D convolutional neural networks that employ spatiotemporal filters on both streams. Although mixing flow with RGB enhances performance, correct optical flow computation is expensive and adds delay to action recognition. In this study, we present a method for training a 3D CNN using RGB frames that replicates the motion stream and, as a result, does not require flow calculation during testing. To begin, in contrast to the SE block, we suggest a channel excitation module (CE module). Experiments have shown that the CE module can improve the feature extraction capabilities of a 3D network and that the effect is superior to the SE block. Second, for action recognition training, we adopt a linear mix of loss based on knowledge distillation and standard cross-entropy loss to effectively leverage appearance and motion information. The Intensified Motion RGB Stream is the stream trained with this combined loss (IMRS). IMRS surpasses RGB or Flow as a single stream; for example, HMDB51 achieves 73.5% accuracy, while RGB and Flow streams score 65.6% and 69.1% accuracy, respectively. Extensive experiments confirm the effectiveness of our proposed method. The comparison with other models proves that our model has good competitiveness in behavior recognition.

Smart IoMT-based segmentation of coronavirus infections using lung CT scans

Computed Tomography (CT) is one of the biomedical imaging modalities which are used to confirm COVID-19 cases and/or to identify infected areas in the lung. Therefore, this article aims at assisting this crucial radiological task by proposing squeeze-and-excitation networks (SENets) within the Internet of medical things (IoMT) framework for automated segmentation of COVID-19 infections in lung CT images. The proposed SE block has been directly integrated with deep residual networks to form Seresnets based on U-Net and LinkNet models. Extensive tests were conducted on a public COVID-19 CT dataset including 20 cases and 1800 + annotated slices to evaluate the segmentation results of our proposed method. The proposed Seresnet models showed a good performance with a Dice score of 0.73, structure similarity index of 0.98, enhanced alignment measure of 0.98, and mean absolute error of 0.06. This study demonstrated a new advanced tool for radiologists to achieve automatic segmentation of the COVID-19 infected areas using CT scans. The main prospect of this research work is deploying our proposed IoMT segmentation framework in the medical diagnosis routine of positive COVID-19 patients.

Auto-segmentation of pancreatic tumor in multi-modal image using transferred DSMask R-CNN network

Pancreatic tumor segmentation is a difficult task due to the high variable shape, small size and hidden position of organs in patients for adaptive radiation therapy plan. To address the problems of limited labeled data, intra-class inconsistency and inter-class indistinction in pancreas tumor segmentation, a transferred DenseSE-Mask R-CNN (TDSMask R-CNN) Network segmentation model using Dense and SE block embedded is proposed in this paper. The multi-scale features strategy is selected to deal with high variability of pancreas and their tumor. The proposed network can learn complementary information from different modes (PET/MR) images respectively by the attention mechanism to get pancreatic tumor regions in different domain. As a result, the irrelevant information for segmenting the tumor area can be suppressed and get low false positives. Furthermore, accurate tumor location from PET image is transferred MRI training model for guide Dense-SE network learning to alleviate the small label samples and reduce network overfitting. Experimental results show that the proposed method achieves average Dice Similarity Coefficient (DSC) of 78.33%, sensitivity (SEN) of 78.56%, and specificity (SPE) of 99.72% on the collected PET/MR data set, which is superior to the existing method of some literatures. This algorithm can improve the accuracy of pancreatic tumor segmentation.

Facial Expression Recognition in the Wild Using Multi-Level Features and Attention Mechanisms

Learning discriminative features is of vital importance for automatic facial expression recognition (FER) in the wild. In this article, we propose a novel Slide-Patch and Whole-Face Attention model with SE blocks (SPWFA-SE), which jointly perceives the discriminative locality characteristics and informative global features of the face for effective FER. Specifically, the well-designed slide patches are proposed to extract local features. Different from the existing methods, our slide patches not only can maintain the information at the edge area of patches, but also do not need to detect facial landmarks. Moreover, to make the model adaptively focus on the distinguishable regions, an attention module is proposed in the patch level to learn the weight of each patch. Furthermore, squeeze-and-excitation blocks are explored in the channel level to learn the weight of each channel. As such, the proposed multi-level feature extraction and attention mechanisms can enhance the representative ability of the learned features. Extensive experiments on five challenging datasets demonstrate that our method can achieve state-of-the-art performance. Cross database experiments on another three databases show the superior generalization performance of our model. Furthermore, complexity analysis results show that our model contains fewer parameters with fast training advantages than other competing models.