Design and Implementation of a Lightweight GoogLeNet Image Classification Model Integrating Depthwise Separable Convolutions

Aiming at the defects of insufficient number of corrosion resistance prediction models and simple feature extraction of B10 copper-nickel alloy, a corrosion resistance prediction model of B10 copper-nickel alloy based on optimized convolutional neural network is proposed. The convolutional neural network architecture for grain boundary image characteristics is proposed to conclude a step-by-step convolution operation by analysing the traditional convolution operation process, and it proves theoretically that such a step-by-step convolution operation can reduce the consumption of additional parameters; learning pool proposed single-channel operation can reduce the loss of feature information; the multi-layer feature fusion learning strategy makes the expressive force of deep network extraction features diversified. The results of multiple experiments demonstrate that the improved algorithm introduced can improve the prediction accuracy of the model in image classification comparing with the traditional convolutional neural network model, and the improved convolutional neural network model can better achieve the prediction of corrosion resistance of B10 copper-nickel alloy.

Aiming at the problem that the detection of bearing fault diagnosis is rarely applied in the research of image classification, a new method based on residual network and attention mechanism is proposed to identify bearing fault diagnosis. One-dimensional vibration signals are transformed into two-dimensional time-frequency images by continuous wavelet transform (CWT), which are input into the model for classification. In order to solve the problem that the traditional convolutional neural network model ignores the low diagnostic accuracy of channel attention and spatial attention due to the loss of important features, the attention mechanism CBAM module is added to make up for the loss of channel features and spatial features in the traditional model. At the same time, the residual network Resnet combined with the attention mechanism can better capture the global information of the time frequency graph, and make up for the defects of the residual network module. The experimental results show that the model has high diagnostic accuracy in rolling bearing fault diagnosis, which proves that the proposed method is effective and feasible.

Antarctic true-color imagery synthesized using multispectral remote sensing data is effective in reflecting sea ice conditions, which is crucial for monitoring. Deep learning has been explored for sea ice extraction, but traditional convolutional neural network models are constrained by a limited perceptual field, making it difficult to obtain global contextual information from remote sensing images. A novel model named GEFU-Net, a modification of U-Net, is presented. The self-established graph reconstruction module is employed to convert features into graph data and construct the adjacency matrix using a global adaptive average similarity threshold. Graph convolutional networks are utilized to aggregate the features at each pixel, enabling the rapid capture of global context, enhancing the semantic richness of the features, and improving the accuracy of sea ice extraction through graph reconstruction. Experimental results using the sea ice dataset of the Ross Sea in the Antarctic, produced by Sentinel-2, demonstrate that our GEFU-Net achieves the best performance compared to other commonly used segmentation models. Specifically, it achieves an accuracy of 97.52%, an Intersection over Union of 95.66%, and an F1-Score of 97.78%. Additionally, fewer model parameters and good inference speed are demonstrated, indicating strong potential for practical ice mapping applications.

Deep learning theory demonstrates its great power in the complicated data processing field. As a prevalent deep learning technology, convolutional neural network (CNN) has achieved some successful applications in the field of fault detection and classification in the complicated industrial systems because of its strong nonlinear feature extraction capability. However, the basic CNN views all the features equally and can not highlight these features which play a more important role in the fault classification procedure. To deal with this problem, this paper proposes an improved CNN model, called multi-head attention CNN (MACNN), to distinguish the importance of different features for better fault classification performance. In the MACNN framework, the raw training data are firstly analyzed by the multiple convolutional layers for feature extraction. Then, the features obtained by the convolution layers are used as the input of the multi-head attention (MA) layers. The introduction of the MA mechanism can help exploit more meaningful information by emphasizing the important features. Lastly, the softmax classifier is utilized to classify the fault patterns. To test the method performance, one case study on the benchmark Tennessee Eastman industrial system is performed and the results show that MACNN gives the better fault classification results than the traditional feedforward neural network (FNN) and CNN models.

Due to the Covid-19 virus millions of people lost their lives and livelihood all the over the world, so quick and accurate diagnosis ofcovid-19 patients is the emergent need of today. Deep learning (DL) techniques, nowadays widely gained much popularity as compared to machine learning methods. Furthermore, DL is well-known in the field of image processing, whereas X-ray pictures are frequently used to diagnose Covid-19 patients. In this article, HyperCNN approach introduced through which higher accuracy can be achieved as compared to other traditional Convolutional Neural Network (CNN). Normally, the tuning of hyper parameters is done manually, which are both costly and time consuming in order to identify the optimum model with the highest accuracy. As per the experiment done, the traditional CNN model provides the accuracy 96.6% whereas the proposed model which is tuned using Bayesian optimization achieves accuracy 98.33%. Hence, it is observed that there is an improvement of 2% in the performance of the proposed model as compared to the traditional CNN model. With this experiment, we claim that automatic hyper parameter optimization is a powerful method for improving the accuracy in transfer learning.KeywordsHyper parameterOptimizationConvolutional neural networkDeep learningCovid-19 diagnosis

Regarding the construction of the “State Grid Cloud” platform, various businesses of the power grid have their own Wed systems. The data between different websites is scattered and the coupling between resources is low, which can easily form the problem of information islands. This paper is oriented to the semi-structured HTML text data in web pages under the State Grid cloud architecture platform, and uses the Python-based Scrapy framework to collect semi-structured power data information from various power business websites. We propose a semi-structured text data classification model based on BIGRU neural network and Bayesian classifier. BIGRU neural network is used to extract text features, Bayesian classifier is used for classification, and the TF-TDF algorithm is used to assign weights to improve the traditional recurrent neural network model with many parameters and long training model time. We use this method to simulate the semi-structured HTML text data of the State Grid, and conduct a comparative experiment with the traditional neural network model. The experimental results show that the classification algorithm can effectively improve the efficiency and accuracy of power semi-structured text data classification.

The inter-class face classification problem is more reasonable than the intra-class classification problem. To address this issue, we have carried out empirical research on classifying Indian people to their geographical regions. This work aimed to construct a computational classification model for classifying Indian regional face images acquired from south and east regions of India, referring to human vision. We have created an Automated Human Intelligence System (AHIS) to evaluate human visual capabilities. Analysis of AHIS response showed that face shape is a discriminative feature among the other facial features. We have developed a modified convolutional neural network to characterize the human vision response to improve face classification accuracy. The proposed model achieved mean F1 and Matthew Correlation Coefficient (MCC) of 0.92 and 0.84, respectively, on the validation set, outperforming the traditional Convolutional Neural Network (CNN). The CNN-Contoured Face (CNN-FC) model is developed to train contoured face images to investigate the influence of face shape. Finally, to cross-validate the accuracy of these models, the traditional CNN model is trained on the same dataset. With an accuracy of 92.98%, the Modified-CNN (M-CNN) model has demonstrated that the proposed method could facilitate the tangible impact in intra-classification problems. A novel Indian regional face dataset is created for supporting this supervised classification work, and it will be available to the research community.

To effectively monitor the nonlinear wear variation of tools during the processing of titanium alloys, this study proposes a hybrid deep neural network fault diagnosis model that integrates the triangulation topology aggregation optimizer (TTAO), convolutional neural network (CNN), bidirectional long short-term memory network (BiLSTM), and attention mechanism (AM). Firstly, vibration signals from the machine tool spindle are acquired and subjected to the wavelet packet transform (WPT) to extract multi-frequency band energy features as model inputs. Then, the CNN and BiLSTM modules capture the features and temporal relationships of the input signals. Finally, introduction of the AM, combined with the TTAO algorithm, automatically extracts deep features, overcoming issues such as local optima and slow convergence in traditional neural networks, thereby enhancing the accuracy and efficiency of tool wear state recognition. The experimental results demonstrate that the proposed model achieves an average accuracy rate of 98.649% in predicting tool wear states, outperforming traditional backpropagation (BP) networks and standard CNN models.

High voltage circuit breakers is an important switchgear of the power system, and 80 percent of fault of high voltage circuit breakers is caused by mechanical failure. Considering a circuit breaker with VS1 type spring actuator as the subject, and the vibration signal under typical mechanical fault is collected. Then the wavelet packet and energy entropy are used to extract the characteristic value. A diagnosis method is proposed based on particle swarm optimization Hopfield neural network. This method to diagnosis fault mode for high voltage circuit breakers is established by analyzing vibration signals of the mechanism. The results show that the accuracy of the method to diagnosis fault mode based on PSO-BP neural network for high voltage circuit breakers is higher than the method of traditional BP neural network model, and the local minimum problem of traditional BP neural network model is prevented by using PSO-BP neural network model. The method of diagnosis fault based on PSO-BP neural network for high voltage circuit breakers is more accurate and feasible compared with traditional BP neural network.

Intelligent fault diagnosis methods based on Transformer variants have shown excellent performance. However, these methods generally model features globally, which means a poor anti-noise ability and greater computational complexity. Therefore, a lightweight framework for bearing fault diagnosis based on depthwise separable convolution (DSC) and sparse attention mechanism named DSC-SparseFormer is proposed. First, a guided-and-sparse self-attention (GSSA) module, which can adaptively adjust the region of interest for features, is proposed to extract features, thereby reducing sensitivity to redundant information and downgrading computational complexity. This is potentially the first time such a methodology has been proposed and applied. Second, discrete cosine transform (DCT) is applied to map time-domain signals into frequency domain to enrich feature space, and a time-frequency coupling attention (TFCA) module is developed to dynamically integrate time-domain and frequency-domain features. Finally, extensive experiments on the HUSTBearing dataset, DHU dataset, and HIT dataset demonstrate that DSC-SparseFormer achieves high-accuracy (HUSTBearing: 99.67%, DHU: 99.94%, and HIT: 99.93%) and lightweight (Params: 0.003 M, FLOPs: 6.07 M) fault diagnosis. The effectiveness of the GSSA and TFCA is validated by ablation studies, and feature visualisation is employed to improve the interpretability of the DSC-SparseFormer.

Aimed to reduce the excessive cost of neural network, this paper proposes a lightweight neural network combining dilated convolution and depthwise separable convolution. Firstly, the dilated convolution is used to expand the receptive field during the convolution process while maintaining the number of convolution parameters, which can extract more high-level global semantic features and improve the classification accuracy of the network. Second, the use of the depthwise separable convolution reduces the network parameters and computational complexity in convolution operations. The experimental results on the CIFAR-10 dataset show that the proposed method improves the classification accuracy of the network while effectively compressing the network size.KeywordsLightweight neural networkDilated convolutionDepthwise separable convolutionClassification accuracyCloud computing

In recent years, Convolutional Neural Networks (CNN) have revealed noteworthy enactment in classification of image jobs. Because of limited availability of annotated brain tumor MRI data, overfitting and lack of generalization are common challenges faced by traditional CNN models. To deal with these challenges, our studies explore the practice of Pre-Trained model of CNN and Transfer learning practices to expand the correctness of cataloguing the brain tumor through the data of MRI. Our experimental results demonstrate that the incorporation of Pre-Trained model of CNN with Transfer learning significantly improves the rain tumor’s classification accuracy on MRI data compared to traditional CNN models trained from scratch. Moreover, the research outcomes highlight potential of transfer learning as an operative approach for improving generalization capabilities of CNN models in medical imaging tasks with limited annotated data. By incorporating pre-trained CNN replicas and transfer erudition techniques, the objective of this studies aims to elevate brain tumor taxonomy accurateness with the assistance of MRI data.

LSGNet: A lightweight convolutional neural network model for tomato disease identification

The detection of epileptic seizures based on the electroencephalogram (EEG) signal is extremely critical in clinical diagnosis and long-term monitoring of patients. Nonetheless, the interpretation process of EEG recordings is time-consuming, subjective, and much relies on the expertise, which shows that precise and automated methods of the process must be developed. The paper contains a detailed comparative evaluation of epileptic seizure detection with the help of various pre-trained convolutional neural network (CNN) models, namely, VGG19, ResNet50, DenseNet121, GoogleNet, MobileNetV2, and Efficient-NetB0, on the CHB-MIT EEG dataset. A new Dual-Domain Feature Extraction Hybrid CNN Ensemble (DDFE-HCNN-E) framework is suggested to improve the performance of detection. The approach uses Continuous Wavelet Transform (CWT) to transform EEG signals to time-frequency representations, which are then useful in learning deep features. The suggested model combines ResNet50 with DenseNet121 by using feature-level fusion in addition to a Convolutional Block Attention Module (CBAM) and a Temporal-Aware Inception Module (TAIM) to identify discriminating spatial, spectral, and temporal features of EEG signals. The experimental findings show that the suggested DDFE-HCNN-E model is better than the traditional pre-trained CNN models with an accuracy of 93.4%, sensitivity of 92.8 %, and specificity of 94.1 %. This result demonstrates that hybrid feature fusion and attention-based temporal modeling are effective in the process of reliable and automated epileptic seizure detection.

Machine learning-based data-driven models have achieved great success since their invention. Nowadays, the artificial neural network (ANN)-based machine learning methods have made great progress than ever before, such as the deep learning and reinforcement learning, etc. In this study, we coupled the ANN with the K-nearest neighbor method to propose a novel hybrid machine learning (HML) hydrological model for flood forecast purpose. The advantage of the proposed model over traditional neural network models is that it can predict discharge continuously without accuracy loss owed to its specially designed model structure. In order to overcome the local minimum issue of the traditional neural network training, a genetic algorithm and Levenberg–Marquardt-based multi-objective training method was also proposed. Real-world applications of the HML hydrological model indicated its satisfactory performance and reliable stability, which enlightened the possibility of further applications of the HML hydrological model in flood forecast problems.