Search

In healthcare (HC), Internet of Things (IoT) integrated cloud computing provides various features and real-time applications. However, owing to the nature of IoT architecture, their types, various modes of communication and the density of data transformed in the network, security is currently a critical issue in the IoT healthcare (IoT-HC) field. This paper proposes a deep learning (DL) model, namely Adaptive Swish-based Deep Multi-Layer Perceptron (ASDMLP) that identifies the intrusions or attacks in the IoT healthcare (IoT-HC) platform. The proposed model starts by clustering the patients' sensor devices in the network using the Probability-based Fuzzy C-Means (PFCM) model. After clustering the devices, the cluster heads (CHs) among the cluster members are selected based on the energy, distance and degree of the sensor devices for aggregating the data sensed by the medical sensor devices. The base station (BS) sends the patient's data collected by the CHs to the cloud server (CS). At the cloud end, the proposed model implements an IDS by applying training of the DL model in publicly available databases. The DL approach first performs preprocessing of the data and then selects optimal features from the dataset using the Opposition and Greedy Levy mutation-based Coyotes Optimization Algorithm (OGCOA). The ASDMLP trains these optimal features for the detection of HC data intrusions. The outcomes confirm that the proposed approach works well on real-time IoT datasets for intrusion detection (ID) without compromising the energy consumption (EC) and lifespan of the network. [ABSTRACT FROM AUTHOR]

The wind power is considered as a potential renewable energy resource which requires less management cost and effort than the others like as tidal, geothermal, etc. However, the natural randomization and volatility aspects of wind in different regions have brought several challenges for efficiently as well as reliably operating the wind-based power supply grid. Thus, it is necessary to have centralized monitoring centers for managing as well as optimizing the performance of wind power farms. Among different management task, wind speed prediction is considered as an important task which directly support for further wind-based power supply resource planning/optimization, hence towards power shortage risk and operating cost reductions. Normally, considering as traditional time-series based prediction problem, most of previous deep learning-based models have demonstrated significant improvement in accuracy performance of wind speed prediction problem. However, most of recurrent neural network (RNN) as well as sequential auto-encoding (AE) based architectures still suffered several limitations related to the capability of sufficient preserving the spatiotemporal and long-range time dependent information of complex time-series based wind datasets. Moreover, previous RNN-based wind speed predictive models also perform poor prediction results within high-complex/noised time-series based wind speed datasets. Thus, in order to overcome these limitations, in this paper we proposed a novel integrated convolutional neural network (CNN)-based spatiotemporal randomization mechanism with transformer-based architecture for wind speed prediction problem, called as: RTrans-WP. Within our RTrans-WP model, we integrated the deep neural encoding component with a randomized CNN learning mechanism to softy align temporal feature within the long-range time-dependent learning context. The utilization of randomized CNN component at the data encoding part also enables to reduce noises and time-series based observation uncertainties which are occurred during the data representation learning and wind speed prediction-driven fine-tuning processes. [ABSTRACT FROM AUTHOR]

Stance detection is the task of classifying user reviews towards a given topic as either supporting, denying, querying, or commenting (SDQC). Most approaches for solving this problem use only the textual features, including the linguistic features and users' vocabulary choice. A few approaches have shown that information from the network structure like graph model can add value, in addition to the textual features, by providing social connections and interactions that may be vital for the stance detection task. In this paper, we present a novel model that combines the text features with the network structure by (1) creating a graph-structure model based on conversational structure towards specific topics and (2) constructing a tree-gated neural network model (TreeGGNN) to capture structure information among reviews. We evaluate our model on four baseline models, which shows that the combination of text and network can achieve an improvement of 2–6% over the state-of-the-art baselines. [ABSTRACT FROM AUTHOR]

For real-time applications, the performance in classifying the movements should be as high as possible, and the computational complexity should be low. This paper focuses on the classification of five upper arm movements which can be provided as a control for human-machine interface (HMI) based applications. The conventional machine learning algorithms are used for classification with both time and frequency domain features, and k-nearest neighbor (KNN) outplay others. To further improve the classification accuracy, pretrained CNN architectures are employed which leads to computational complexity and memory requirements. To overcome this, the deep convolutional neural network (CNN) model is introduced with three convolutional layers. To further improve the performance which is the key idea behind real-time applications, a hybrid CNN-KNN model is proposed. Even though the performance is high, the computation costs of the hybrid method are more. Minimum redundancy maximum relevance (mRMR), a feature selection method makes an effort to reduce feature dimensions. As a result, better performance is achieved by our proposed method CNN-KNN with mRMR which reduces computational complexity and memory requirement with a mean prediction accuracy of about 99.05±0.25% with 100 features. [ABSTRACT FROM AUTHOR]

Protein Remote Homology and fold Recognition (PRHR) is the most crucial task to predict the protein patterns. To achieve this task, Sequence-Order Frequency Matrix-Sampling and Deep learning with Smith-Waterman (SOFM-SDSW) were designed using large-scale Protein Sequences (PSs), which take more time to determine the high-dimensional attributes. Also, it was ineffective since the SW was only applied for local alignment, which cannot find the most matches between the PSs. Hence, in this manuscript, a rapid semi-global alignment algorithm called SOFM-SD-GlobalSW (SOFM-SDGSW) is proposed that facilitates the affine-gap scoring and uses sequence similarity to align the PSs. The major aim of this paper is to enhance the alignment of SW algorithm in both locally and globally for PRHR. In this algorithm, the Maximal Exact Matches (MEMs) are initially obtained by the bit-level parallelism rather than to align the individual characters. After that, a subgroup of MEMs is obtained to determine the global Alignment Score (AS) using the new adaptive programming scheme. Also, the SW local alignment scheme is used to determine the local AS. Then, both local and global ASs are combined to produce a final AS. Further, this resultant AS is considered to train the Support Vector Machine (SVM) classifier to recognize the PRH and folds. Finally, the test results reveal the SOFM-SDGSW algorithm on SCOP 1.53, SCOP 1.67 and Superfamily databases attains an ROC of 0.97, 0.941 and 0.938, respectively, as well as, an ROC50 of 0.819, 0.846 and 0.86, respectively compared to the conventional PRHR algorithms. [ABSTRACT FROM AUTHOR]

The study examines various studies on Named Entity Recognition (NER) and Part of Speech (POS) tagging for the Urdu language conducted by academics and researchers. POS and NER tagging for Urdu still faces obstacles in terms of increasing accuracy while lowering false-positive rates and labelling unknown terms, despite the efforts of numerous researchers. In addition, ambiguity exists when tagging terms with different contextual meanings within a sentence. Due to the fact that Urdu is an inflectional, derivational, morphologically rich, and context-sensitive language, the existing models, such as Linguistic rule application, N-gram Markov model, Tree Tagger, random forest (RF) tagger, etc., were unable to produce accurate experimental results on Urdu language data. The significance of this study is that it fills a gap in the literature concerning the lack of POS and NER tagging for the Urdu language. For Urdu POS and NER tagging, we propose a deep learning model with a well-balanced set of language-independent features as well as a survey of important Urdu POS/NER techniques. In addition, this is the first study to use residual biDirectional residual Long short-term memory (residual biLSTM) architecture trained on the Urmono dataset in conjunction with the randomly initialised word2vec, fastText and mBERT embeddings are utilised to generate word or character vectors.For each experiment, the paper also employs the evaluation methods of Macro-F1, precision, precision, and recall. The proposed method with mbert embedding as word vectors provides best results of F1 score for POS and NER at 91.11% and 99.11% respectively. Also, the accuracy, precision and recall for POS are reported at 94.85%, 91.79% and 90.77%. Similarly, the accuracy, precision and recall for NER of the proposed model are reported at 99.77%, 98.78% and 99.45% respectively, which are higher than baseline models. [ABSTRACT FROM AUTHOR]

The surge of renewable energy systems can lead to increasing incidents that negatively impact economics and society, rendering incident detection paramount to understand the mechanism and range of those impacts. In this paper, a deep learning framework is proposed to detect renewable energy incidents from news articles containing accidents in various renewable energy systems. The pre-trained language models like Bidirectional Encoder Representations from Transformers (BERT) and word2vec are utilized to represent textual inputs, which are trained by the Text Convolutional Neural Networks (TCNNs) and Text Recurrent Neural Networks. Two types of classifiers for incident detection are trained and tested in this paper, one is a binary classifier for detecting the existence of an incident, the other is a multi-label classifier for identifying different incident attributes such as causal-effects and consequences, etc. The proposed incident detection framework is implemented on a hand-annotated dataset with 5 190 records. The results show that the proposed framework performs well on both the incident existence detection task (F1-score 91.4%) and the incident attributes identification task (micro F1-score 81.7%). It is also shown that the BERT-based TCNNs are effective and robust in detecting renewable energy incidents from large-scale textual materials. [ABSTRACT FROM AUTHOR]

In this paper, we address the challenge of producing fully automated land cover estimates from satellite imagery through Deep Learning algorithms. We developed our system according to a tile-based, classify-and-count design. To implement the classification engine of the system, we adopted a cutting-edge Convolutional Neural Network model named Inception-V3, which we customized and trained for scene classification on the EuroSAT dataset. We tested and validated our system on two Sentinel-2 images representing quite different Italian territories (with an area of 751 km 2 and 443 km 2 , respectively). Because no genuine ground-truth is available for the land cover of these sub-regional territories, we built a pseudo ground-truth by integrating land cover information from flagship European projects CORINE and LUCAS. A critical and careful analysis shows that our automatic land cover estimates are in good agreement with the pseudo ground-truth and offers extensive evidence of the remarkable segmentation ability of our system. The limits of our approach are also critically discussed in the paper and possible countermeasures are illustrated. When compared with traditional projects like CORINE and LUCAS, our automatic land cover estimation system exhibits three fundamental advantages: it can dramatically reduce production costs; it can allow delivering very timely and frequent land cover statistics; it can enable land cover estimation for very small territorial areas, well beyond the NUTS-2 level. As an additional outcome of land cover estimation, our system also automatically generates moderate resolution land cover maps that might be used in cartography projects as an agile first-level tool for map update or change detection purposes. [ABSTRACT FROM AUTHOR]

COVID-19 is a contagious respiratory illness that can be passed from person to person. Because it affects the lungs, damages blood arteries, and causes cardiac problems, COVID-19 must be diagnosed quickly. The reverse transcriptase polymerase chain reaction (RT-PCR) is a method for detecting COVID-19, but it is time consuming and labor expensive, as well as putting the person collecting the sample in danger. As a result, clinicians prefer to use CT scan and Xray images. COVID-19 classification can be done manually, however AI makes the process go faster. AI approaches include image processing, machine learning, and deep learning. An AI model is required to diagnose COVID-19, and a dataset is necessary to train that model. A dataset consists of the information from which the model is trained. This paper consists of the review of different image processing, machine learning and deep learning papers proposed by different researchers. As well as models based on deep learning and pretrained model using gradient boosting algorithm The goal of this paper is to provide information for future researchers to work with. [ABSTRACT FROM AUTHOR]

Federated learning (FL) refers to a system in which a central aggregator coordinates the efforts of several clients to solve the issues of machine learning. This setting allows the training data to be dispersed in order to protect the privacy of each device. This paper provides an overview of federated learning systems, with a focus on healthcare. FL is reviewed in terms of its frameworks, architectures and applications. It is shown here that FL solves the preceding issues with a shared global deep learning (DL) model via a central aggregator server. Inspired by the rapid growth of FL research, this paper examines recent developments and provides a comprehensive list of unresolved issues. Several privacy methods including secure multiparty computation, homomorphic encryption, differential privacy and stochastic gradient descent are described in the context of FL. Moreover, a review is provided for different classes of FL such as horizontal and vertical FL and federated transfer learning. FL has applications in wireless communication, service recommendation, intelligent medical diagnosis system and healthcare, which we review in this paper. We also present a comprehensive review of existing FL challenges for example privacy protection, communication cost, systems heterogeneity, unreliable model upload, followed by future research directions. [ABSTRACT FROM AUTHOR]

Taking Uyghur character recognition as an example, this paper use a method of image segmentation which combines traditional methods with CNN, and transforms and implements the Uyghur character on intelligent devices. Firstly, after analyzing the application of several general segmentation methods, this paper finds some shortcomings but also some ideas for Uighur segmentation. Then, starting from the characteristics of Uyghur language, such as structure, word formation and input habits, the author studies the idea of Uyghur adhesive language segmentation from the perspective of language characteristics, puts forward the basic algorithm of Uyghur symbol segmentation, and applies the Uyghur character adhesion segmentation based on minimum spanning tree and multi-queue primitive merging model to improve the segmentation efficiency. In addition, in order to solve the limitations of the traditional handwriting recognition framework of "preprocessing + feature extraction + classifier", this paper puts forward a new solution of Uyghur handwriting recognition technology combining prior domain knowledge with CNN, constructs a Uyghur handwriting recognition model on the basis of CNN and random elastic deformation, and also improves the recognition rate of the system by combining domain knowledge with CNN. According to the experimental results, we can conclude that the mixed method proposed in this paper can effectively break the technical bottleneck of traditional methods, thus improving the efficiency of segmentation and recognition. [ABSTRACT FROM AUTHOR]

As an advanced machine vision task, traffic sign recognition is of great significance to the safe driving of autonomous vehicles. Haze has seriously affected the performance of traffic sign recognition. This paper proposes a dehazing network, including multi-scale residual blocks, which significantly affects the recognition of traffic signs in hazy weather. First, we introduce the idea of residual learning, design the end-to-end multi-scale feature information fusion method. Secondly, the study used subjective visual effects and objective evaluation metrics such as Visibility Index (VI) and Realness Index (RI) based on the characteristics of the real-world environment to compare various traditional dehazing and deep learning dehazing method with good performance. Finally, this paper combines image dehazing and traffic sign recognition, using the algorithm of this paper to dehaze the traffic sign images under real-world hazy weather. The experiments show that the algorithm in this paper can improve the performance of traffic sign recognition in hazy weather and fulfil the requirements of real-time image processing. It also proves the effectiveness of the reformulated atmospheric scattering model for the dehazing of traffic sign images. [ABSTRACT FROM AUTHOR]

The identification and classification of plant diseases is of great significance to ecological protection and deep learning methods have made a great of progress in the common plant diseases identification for specific plant. While faced with the same plant disease of other plants, due to the insufficient or low quality training data, current deep learning methods will be difficult to identify the diseases effectively and accurately. Inspired by the advantages of GAN in dataset expansion, we propose the CycleGAN based confusion model in this paper. In this paper, GAN framework is improved by adding noise label and learn together during training stage, which migrates the data of common plant diseases to the plants with insufficient or low quality data. In order to evaluate the quality of the migrated training dataset among different GAN approaches, we introduce the quality indicators of the migration images such as MMD, FID, EMD etc. We compare our model with other GANs model, and the experimental results show that the proposed model obtains better results in the migration process, which make it more effective for the identification of cross species plant diseases. [ABSTRACT FROM AUTHOR]

Arabic language has a set of sound letters called diacritics, these diacritics play an essential role in the meaning of words and their articulations. The change in some diacritics leads to a change in the context of the sentence. However, the existence of these letters in the corpus transcription affects the accuracy of speech recognition. In this paper, we investigate the effect of diactrics on the Arabic speech recognition based end-to-end deep learning. The applied end-to-end approach includes CNN-LSTM and attention-based technique presented in the state-of-the-art framework namely, Espresso using Pytorch. In addition, and to the best of our knowledge, the approach of CNN-LSTM with attention-based has not been used in the task of Arabic Automatic speech recognition (ASR). To fill this gap, this paper proposes a new approach based on CNN-LSTM with attention based method for Arabic ASR. The language model in this approach is trained using RNN-LM and LSTM-LM and based on nondiacritized transcription of the speech corpus. The Standard Arabic Single Speaker Corpus (SASSC), after omitting the diacritics, is used to train and test the deep learning model. Experimental results show that the removal of diacritics decreased out-of-vocabulary and perplexity of the language model. In addition, the word error rate (WER) is significantly improved when compared to diacritized data. The achieved average reduction in WER is 13.52%. [ABSTRACT FROM AUTHOR]

Internet of Vehicles (IoV) presents a new generation of vehicular communications with limited computation offloading, energy and memory resources with 5G/6G technologies that have grown enormously and are being used in wide variety of Intelligent Transportation Systems (ITS). Due to the limited battery power in smart vehicles, the concept of energy consumption is one of the main and critical challenges of the IoV environments. Optimizing resource management strategies for improving the energy consumption using AI-based methods is one of important solutions in the IoV environments. There are various machine learning algorithms for selecting optimal solutions for energy-efficient resource management strategies. This paper presents the existing energy-aware resource management strategies for the IoV case studies, and performs a comparative analysis among their applied AI-based methods and machine learning algorithms. This analysis presents a technical and deeper understanding of the technical aspects of existing machine learning and AI-based algorithms that will be helpful in design of new hybrid AI approaches for optimizing resource management strategies with reducing their energy consumption. [ABSTRACT FROM AUTHOR]

This paper shows the capabilities offered by an integrated neural-logic multi-agent system (MAS). Our case study encompasses logical agents and a deep learning (DL) component, to devise a system specialised in monitoring flood events for civil protection purposes. More precisely, we describe a prototypical framework consisting of a set of intelligent agents, which perform various tasks and communicate with each other to efficiently generate alerts during flood crisis events. Alerts are only delivered when at least two separates sources agree on an event on the same zone, i.e. aerial images and severe weather reports. Images are segmented by a neural network trained over eight classes of topographical entities. The resulting mask is analysed by a Logic Image Descriptor (LID) which then submit the perception to a logical agent. [ABSTRACT FROM AUTHOR]

The edge features of peanut grain image are not considered in peanut grain integrity detection, and there are many noises, resulting in low accuracy of peanut grain integrity detection and poor effect of image edge noise removal. Therefore, this paper designs a peanut seed integrity detection method based on deep learning convolution neural network. The edge contour of peanut grain image is closed, the edge contour curve characteristics of peanut grain image are extracted, the noise area of peanut grain edge image is determined according to the filter window of Gaussian kernel function, and the wavelet descriptor in contour description operator is used to reduce the edge noise of peanut grain image. Input the image into the convolution neural network, update the weight by gradient descent method, construct the peanut seed integrity detection model, output the peanut seed integrity detection results, and realize the peanut seed integrity detection. The experimental results show that the proposed detection method can improve the accuracy of peanut grain detection and has certain feasibility. [ABSTRACT FROM AUTHOR]

Generative Adversarial Networks (GANs) has achieved great success in computer vision like Image Inpainting, Image Super-Resolution. Many researchers apply it to improve the effectiveness of recommendation system. However, GANs-based methods obtain users' preferences using a single Neural Network framework in generative model, which may not be fully mined. Furthermore, most GANs-based algorithms adopt cross-entropy loss to get pair-wise bias, but these methods don't reveal global data distribution loss when data are sparse. Those problems will influence the performance of the algorithm and result in poor accuracy. To address these problems, we introduce Wide & Deep Generative Adversarial Networks for Recommendation System (a.k.a W & DGAN) in this paper. On the one hand, we employ Wide & Deep Learning as a generative model capable of extracting both explicit and implicit information of user preferences. Furthermore, we combine Cross-Entropy loss in G with Wasserstein loss in D to get data distribution, then, the joint loss will be to receive the training information feedback from data distribution. Empirical results on three public benchmarks show that W&DGAN significantly outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Vehicle Re-Identification (Re-ID) aims to discover and match target vehicles in different cameras of road surveillance. The high similarity between vehicle appearances and the dramatic variations in viewpoints and illumination cause great challenges for vehicle Re-ID. Meanwhile, in safety supervision and intelligent traffic systems, one needs a quick efficient method of identifying target vehicles. In this paper, we propose a Multi-Attention Guided Feature Enhancement Network (MAFEN) to extract robust vehicle appearance features. Specifically, the Fusing Spatial-Channel information multi-receptive fields Feature Enhancement module (FSCFE) is first proposed to aggregate richer and more representative multi-receptive fields features at different receptive fields sizes. It also learned the spatial structure information and channel dependencies of the multi-receptive fields features and embedded them to enhance the feature. Then, we construct the Spatial Attention-Guided Adaptive Feature Erasure (SAAFE) module, which uses spatial attention to erase the most distinguishing features. The network's attention is shifted to potentially salient features to strengthen the ability of the network to extract salient features. In addition, a multi-loss knowledge distillation (MLKD) method using MAFEN as a teacher network is designed to improve computational efficiency. It uses multiple loss functions to jointly supervise the student network. Experimental results on three public datasets demonstrate the merits of the proposed method over the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The use of command and control (C2) servers in cyberattacks has risen considerably, attackers frequently employ the domain generated algorithm (DGA) technique to conceal their C2 servers. Various machine learning models have been suggested for binary identification of domain names as either benign or DGA domain. The Existing techniques are inefficient and have real-time detection issues and are also very data hypersensitive, therefore, they can be circumvented by the attackers. The main problem this article addresses is how to automatically detect DGA in a way that does not rely solely on reverse engineering, not strongly affected by data size, and allows detection of this DGA in real time. This paper presents DTFS-DGA model that combine neural networks models with traditional machine learning models and maintains its performance even if the data size changes to detect DGA in real time. The model uses 15 linguistics and networks features with the features extracted by long short-term memory and convolutional neural network to classify domain names using random forest and support vector machines. The comprehensive experimental findings confirm the suggested model's accuracy. To be precise, the model achieve an average accuracy of 99.8 % for the classification. [ABSTRACT FROM AUTHOR]

Recent years have witnessed a rapid advance in training and testing synthetic data through deep learning networks for the annotation of synthetic data that can be automatically marked. However, a domain discrepancy still exists between synthetic data and real data. In this paper, we address the domain discrepancy issue from three aspects: 1) We design a synthetic image generator with automatically labeled based on 3D scenes. 2) A novel adversarial domain adaptation model is proposed to learn robust intermediate representation free of distractors to improve the transfer performance. 3) We construct a distractor-invariant network and adopt the sample transferability strategy on global-local levels, respectively, to mitigate the cross-domain gap. Additional exploratory experiments demonstrate that the proposed model achieves large performance margins, which show significant advance over the other state-of-the-art models, performing a promotion of 10%–15% mAP on various domain adaptation scenarios. [ABSTRACT FROM AUTHOR]

Insulator determines the insulation level and power supply reliability of transmission lines. The traditional operation and maintenance method of insulators has a large workload. This paper presents an insulator recognition and fault diagnosis system based on image recognition and machine learning. Firstly, the composite insulators in complex backgrounds have been identified by Faster RCNN algorithm, which helps to extract the image of insulators by drone shot. Then, the cracking of umbrella skirts has been carried out by means of image processing. Also, the contamination of composite insulator umbrella skirts is also identified. The appropriate feature quantity by Fisher's discrimination is recommended, and the insulator contamination level has been identified by S component. Lastly, analyzing the infrared image of the insulator provides the basis for the normalization of insulator temperature rise detection results in different environments. Insulator image autonomous recognition and defect intelligent detection, which is based on deep reinforcement learning, is helpful for the operation, maintenance and repair of line insulators. [ABSTRACT FROM AUTHOR]

Speech denoising refers to people finding useful information in noise-affected speech data and processing it. This technology is an important part of speech signal processing technology. With the continuous progress of science and technology, its application in various fields is gradually improving, among which voice noise reduction technology has also been well developed. Especially in recent years, with the rapid development of artificial intelligence science and technology, anti-network speech denoising technology based on deep learning has been derived. Compared with traditional speech denoising techniques, it has better performance in resolving speech data distortion and denoising results. However, in practical applications, this method still has some shortcomings, such as: some unique noises are not removed, and the effective information extraction is distorted under the background of high-intensity noise. In response to these outstanding problems, this research is based on the anti-network speech denoising technology, combined with the Wiener filter speech denoising algorithm, and applies artificial intelligence technology to network security technology management. In this paper, two algorithms, COVL and SSNR, are used to compare the original input speech data signal with noise and the speech data signal preprocessed by Wiener filtering. The additional information is input into the generation network and the discriminant network, and the additional information is used to guide the data generation. The least squares loss function of the binary coding algorithm is used to replace the cross entropy loss function, so that the original discriminant network loss function is not changed, which improves the accuracy of discrimination. This improved technology is superior in terms of speed, productivity and improved accuracy. While building the system, mature network technologies and equipment applications are fully applied, and speech de-noising technology is used for network security situation assessment and prediction simulation tests. The simulation test results show that the network security situation assessment is reasonable and effective, and the prediction is accurate in real time. [ABSTRACT FROM AUTHOR]

This paper studies the robot-written character identification problem under an end-to-end semi-supervised deep learning framework consisting of semi-supervised learning and deep learning modules. The learning framework allows a deep neural network to be trained on labeled and pseudo-labeled samples where pseudo-labeled samples refer to the samples with labels predicted by the semi-supervised learning module. Moreover, to guarantee the feasibility of the learning framework, a two-stage strategy is proposed for training the deep neural network. Specifically, the two-stage training strategy adopts pseudo-labeled samples firstly to train a deep neural network, then the deep neural network is refined using labeled samples one more time. As a result, more samples can be used for training a deep neural network, which is significant to the performance improvement of a deep neural network in the case of inadequate labeled samples. More importantly, the deep neural networks trained under the proposed learning framework perform better than the famous deep neural networks in a robot-written character identification experiment. [ABSTRACT FROM AUTHOR]

Cervical cancer is the most frequent and fatal malignancy among women worldwide. If this tumor is detected and treated early enough, the complications it causes can be minimized. Deep learning demonstrated significant promise when imposed on biomedical difficulties such as medical image processing and disease prognostication. Therefore, in this paper, an automatic cervical cell classification approach named IR-PapNet is developed based on Inception-ResNet which is an optimized version of Inception. The learning model's conventional ReLu activation is replaced with the parametric-rectified linear unit (PReLu) to overcome the nullification of negative values and dying ReLu. Finally, the model loss function is minimized with the SGD optimization model by modifying the attributes of the neural network. Furthermore, we present a simple but efficient noise removal technique called 2D-Discrete Wavelet Transform (2D-DWT) algorithm for enhancing image quality. Experimental results show that this model can achieve a top-1 average identification accuracy of 99.8% on the pap smear cervical Herlev datasets, which verifies its satisfactory performance. The restructured Inception-ResNet network model can obtain significant improvements over most of the state-of-the-art models in 2-class classification, and it achieves a high learning rate without experiencing dead nodes. [ABSTRACT FROM AUTHOR]

This paper presents the design of a grid connected hybrid system using modified Z source converter, bidirectional converter and battery storage system. The input sources for the proposed system are fed from solar and wind power systems. A modified high gain switched Z source converter is designed for supplying constant DC power to the DC-link of the inverter. A hybrid deep learning (HDL) algorithm (CNN-BiLSTM) is proposed for predicting the output power from the hybrid systems. The HDL method and the PI controller generates pulses to the proposed system. The superiority of the proposed hybrid DL method is compared with the conventional DL methods like CNN, LSTM, BiLSTM methods and the performance of the hybrid system is validated. A closed loop control framework is implemented for the proposed grid integrated hybrid system and its performance is observed by implementing the PI, Fuzzy and ANN controllers. A 1.5Kw hybrid system is designed in MATLAB/SIMULINK software and the results are validated. A prototype of the proposed system is developed in the laboratory and experimental results are obtained from it. From the simulation and experimental results, it is observed that the ANN controller with SVPWM (Space vector Pulse width Modulation) gives a THD (Total harmonic distortion) of 2.2% which is within the IEEE 519 standard. Therefore, from the results it is identified that the ANN-SVPWM method injects less harmonic currents into the grid than the other two controllers. [ABSTRACT FROM AUTHOR]

Energy or load disaggregation, as one essential part of non-intrusive load monitoring (NILM), is an efficient way to separate the consumption information of target appliances from the whole consumption data, and can accordingly help to regulate people's energy consumption behaviors. However, the consumptions of the target appliances are usually affected by the variance of the opening time, working condition and user interference, so it is a difficult task to realize precise disaggregation. To further improve the energy disaggregation accuracy, this paper proposes a new parallel disaggregation strategy with two subnets for the energy consumption disaggregation of the target appliances in the residential buildings. In the proposed strategy, the parallel disaggregation network contains a long-term disaggregation network and a short-term disaggregation network, which can automatically and respectively learn the long-term trend features and short-term dynamic characteristics of the electrical appliances. This parallel structure can make full use of the advantages of different methods in feature extraction, so as to model the appliance features more comprehensively. To better extract the long-term and short-term features, in the long-term disaggregation subnet, we propose the double branch bi-directional temporal convolution network (DBB-TCN) which has a wider receptive field than the traditional temporal convolution networks (TCN), while in the short-term disaggregation subnet, we adopt the convolution auto-encoder to learn the short-term characteristics of the target appliances. Finally, detailed experiments and comparisons are made with two real-world datasets. Experimental results verified that the proposed parallel disaggregation method performs better than the existing methods under various evaluation criteria. [ABSTRACT FROM AUTHOR]

Many real-world knowledge graphs are complex and keep evolving over time. Inferring missing facts in temporal knowledge graphs is a fundamental and challenging task. Previous studies focus on link prediction in static knowledge graphs which hardly extracts the temporal features effectively. In this paper, we propose a novel deep learning model, namely KBGAT-BiLSTM, which is capable of solving long-term predict problems and is suitable for temporal knowledge graph with complex structures. First, we adapt the Graph Attention Network (GAT) to learn the structural features of knowledge graph. Then we utilize the Bidirectional Long Short-Term Memory Networks (BiLSTM) to learn the temporal features and obtain the low-dimensional embeddings of entities and relations. Finally, we employ a scoring function for link prediction in temporal knowledge graphs. Through extensive experiments on YAGO, WIKI, and ICEWS18 datasets, we demonstrate the effectiveness of our model, compare the performance of our model with several different state-of-the-art methods and further analyze the properties of the proposed method. [ABSTRACT FROM AUTHOR]

Keyphrases are important phrases that represent the theme of a document. With the help of keyphrases people can quickly find useful information from massive data. Traditional statistic-based methods for keyphrase extraction only make use of the statistical features of the words and ignore the semantic relationship between words. Recently, the emerging methods based on deep neural network extract keyphrases by capturing the semantic contextual information without considering the statistical features. In this paper, we propose a new keyphrase extraction method based on the neural network architecture composing of deep and wide learning parts. In the deep learning part, BERT (Bidirectional Encoder Representation from Transformers) and Bi-LSTM (Bidirectional Long Short-Term Memory) models are used to capture the contextual semantic information from the word sequence while in the wide learning part several important statistical features are considered to jointly train the keyphrase extraction model. The experimental results on two public datasets show that the performance of our proposed model is better than eight commonly baseline keyphrase extraction methods. [ABSTRACT FROM AUTHOR]

Representing features at multiple scales is of great significance for hyperspectral image classification. However, the most existing methods improve the feature representation ability by extracting features with different resolutions. Moreover, the existing attention methods have not taken full advantage of the HSI data, and their receptive field sizes of artificial neurons in each layer are identical, while in neuroscience, the receptive field sizes of visual cortical neurons adapt to the neural stimulation. Therefore, in this paper, we propose a Res2Net with spectral-spatial and channel attention (SSCAR2N) for hyperspectral image classification. To effectively extract multi-scale features of HSI image at a more granular level while ensuring a small amount of calculation and low parameter redundancy, the Res2Net block is adopted. To further recalibrate the features from spectral, spatial and channel dimensions simultaneously, we propose a visual threefold (spectral, spatial and channel) attention mechanism, where a dynamic neuron selection mechanism that allows each neuron to adaptively adjust the size of its receptive fields based on the multiple scales of the input information is designed. The comparison experiments on three benchmark hyperspectral image data sets demonstrate that the proposed SSCAR2N outperforms several state-of-the-art deep learning based HSI classification methods. [ABSTRACT FROM AUTHOR]

Diabetic Retinopathy (DR) is a disease that damages the retina of the human eye due to diabetic complications, resulting in a loss of vision. Blindness may be avoided If the DR disease is detected at an early stage. Unfortunately, DR is irreversible process, however, early detection and treatment of DR can significantly reduce the risk of vision loss. The manual diagnosis done by ophthalmologists on DR retina fundus images is time consuming, and error prone process. Nowadays, machine learning and deep learning have become one of the most effective approaches, which have even surpassed the human performance as well as performance of traditional image processing-based algorithms and other computer aided diagnosis systems in the analysis and classification of medical images. This paper addressed and evaluated the various recent state-of-the-art methodologies that have been used for detection and classification of Diabetic Retinopathy disease using machine learning and deep learning approaches in the past decade. Furthermore, this study also provides the authors observation and performance evaluation of available research using several parameters, such as accuracy, disease status, and sensitivity. Finally, we conclude with limitations, remedies, and future directions in DR detection. In addition, various challenging issues that need further study are also discussed. [ABSTRACT FROM AUTHOR]

With the growth in vehicular traffic, there is a greater risk of road accidents. Over speeding, intoxicated driving, driver distractions, red-light runners, ignoring safety equipment such as seat belts and helmets, non-adherence to lane driving, and improper overtaking are the leading causes of accidents. Speed violation, in particular, has a significant influence on today's transportation. Also, detecting this speed violation and punishing this violator are more time-consuming tasks. For that reason, a novel automatic speed violation detection in traffic based on Deep learning is proposed in this paper. This proposed method is separated into two working modules: object detection and license plate recognition. The object detection module uses the most efficient PP YOLO neural networks. It utilizes open ALPR (Automatic License Plate Recognition) for the vehicle's number plate identification, which passes the traffic above maximum speed. With the number plate details, the authorities can take action against the rule violator with less time and effort. The simulation results show that the proposed automatic speed violation detection system also has an accuracy rate of 98.8% for speed violation detection and 99.3% for license plate number identification, demonstrating that the approach described in this work has a higher performance in terms of accuracy. Furthermore, the proposed technique was compared to recent existing results. [ABSTRACT FROM AUTHOR]

Recognition of finger vein patterns is essential technique that analyses the finger vein patterns to enable accurate authentication of an individual. A proper, accurate and quick learning of patterns is essentially required for improving the classification pattern. It is essential in developing an intelligent algorithm to effectively study and classify the patterns. In this paper, we develop an improved deep learning hybrid model for feature extraction and classification. A dimensional reduction deep neural network (DR-DNN) model has included a dimensional reduction model for extracting the essential features by reducing the dimensionality of feature datasets. A convolutional neural network (CNN) helps in classifying the benign vein patterns from the malignant vein patterns. The effectiveness is compared against existing deep learning classifiers to measure how effective the deep learning model is used for classifying finger vein patterns for biometric authentication. The results shows that the proposed method achieves an accuracy rate of 97.16% for the proposed method, where the other existing methods including CNN, Recurrent Neural Network (RNN) and Deep Neural Nets (DNN) has an accuracy rate of 86%, 80.66% and 88.31%, respectively. [ABSTRACT FROM AUTHOR]

With the rapid industrialization and urbanization worldwide, air quality levels are deteriorating at an unprecedented rate and posing a substantial threat to humans and the environment. This brings the concern to effectively monitor and forecast air quality levels in real-time. Conventional air quality monitoring stations are built based on centralized architectures involving high latency, communication technologies demanding high power, sensors involving high costs and decision making with moderate accuracy. To address the limitations of the existing systems, we propose a smart and distinct Air Quality Monitoring and Forecasting system embracing Fog Computing with IoT and Deep Learning (DL). The system is a three-layered architecture with the Sensing layer first, Fog Computing layer in between, and Cloud Computing layer at the end. Fog Computing is a powerful new generation paradigm that brings storage, computation, and networking at the edge of the IoT network and reduce network latency. A DL based BiLSTM (Bidirectional Long Short-Term Memory) model is deployed in the Fog Computing layer. The proposed system aims at real-time monitoring and accurate air quality forecasting to support decision making and aid timely prevention and control of pollutant emissions by alerting the stakeholders when a dangerous Air Quality Index (AQI) is expected. Experimental results show that the BiLSTM model has a better predictive performance considering the meteorological parameters than the baseline models in terms of MAE and RMSE. A proof of concept realizing the proposed system is elaborated in the paper. [ABSTRACT FROM AUTHOR]

Due to the wide range of diseases and imaging modalities, a retrieving system is a challenging task to access the corresponding clinical cases from a large medical repository on time. Several computer-aided systems (CADx) are developed to recognize medical imaging modalities (MIM) based on various standard machine learning (SML) and advanced deep learning (DL) algorithms. Pre-trained models like convolutional neural networks (CNN) are used in the past as a transfer learning (TL) architecture. However, it is a challenging task to use these pre-trained models for some unseen datasets with a different domain of features. To classify different medical images, the relevant features with a robust classifier are needed and still, it is unsolved task due to MIM-based features. In this paper, a hybrid MIM-based classification system is developed by integrating the pre-trained VGG-19 and ResNet34 models into the original CNN model. Next, the MIM-DTL model is fine-tuned by updating the weights of new layers as well as weights of original CNN layers. The performance of MIM-DTL is compared with state-of-the-art systems based on cancer imaging archive (TCIA), Kvasir and lower extremity radiographs (LERA) datasets in terms of statistical measures such as accuracy (ACC), sensitivity (SE) and specificity (SP). On average, the MIM-DTL model achieved 99% of ACC, SE of 97.5% and SP of 98% along with smaller epochs compare to other TL. The experimental results show that the MIM-DTL model is outperformed to recognize medical imaging modalities and helps the healthcare experts to identify relevant diseases. [ABSTRACT FROM AUTHOR]

Human emotion recognition with the evaluation of speech signals is an emerging topic in recent decades. Emotion recognition through speech signals is relatively confusing because of the speaking style, voice quality, cultural background of the speaker, environment, etc. Even though numerous signal processing methods and frameworks exists to detect and characterize the speech signal's emotions, they do not attain the full speech emotion recognition (SER) accuracy and success rate. This paper proposes a novel algorithm, namely the deep ganitrus algorithm (DGA), to perceive the various categories of emotions from the input speech signal for better accuracy. DGA combines independent component analysis with fisher criterion for feature extraction and deep belief network with wake sleep for emotion classification. This algorithm is inspired by the elaeocarpus ganitrus (rudraksha seed), which has 1 to 21 lines. The single line bead is rarest to find, analogously finding a single emotion from the speech signal is also complex. The proposed DGA is experimentally verified on the Berlin database. Finally, the evaluation results were compared with the existing framework, and the test result accomplishes better recognition accuracy when compared with all other current algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we propose CryptMed, a system framework that enables medical service providers to offer secure, lightweight, and accurate medical diagnostic service to their customers via an execution of neural network inference in the ciphertext domain. CryptMed ensures the privacy of both parties with cryptographic guarantees. Our technical contributions include: 1) presenting a secret sharing based inference protocol that can well cope with the commonly-used linear and non-linear NN layers; 2) devising optimized secure comparison function that can efficiently support comparison-based activation functions in NN architectures; 3) constructing a suite of secure smooth functions built on precise approximation approaches for accurate medical diagnoses. We evaluate CryptMed on 6 neural network architectures across a wide range of non-linear activation functions over two benchmark and four real-world medical datasets. We comprehensively compare our system with prior art in terms of end-to-end service workload and prediction accuracy. Our empirical results demonstrate that CryptMed achieves up to respectively 413x, 19x, and 43x bandwidth savings for MNIST, CIFAR-10, and medical applications compared with prior art. For the smooth activation based inference, the best choice of our proposed approximations preserve the precision of original functions, with less than 1.2% accuracy loss and could enhance the precision due to the newly introduced activation function family. [ABSTRACT FROM AUTHOR]

In the deep learning-based video action recognitio, the function of the neural network is to acquire spatial information, motion information, and the associated information of the above two kinds of information over an uneven time span. This paper puts forward a network extracting video sequence semantic information based on deep integration of local Spatial-Temporal information. The network uses 2D Convolutional Neural Network (2DCNN) and Multi Spatial-Temporal scale 3D Convolutional Neural Network (MST_3DCNN) respectively to extract spatial information and motion information. Spatial information and motion information of the same time quantum receive 3D convolutional integration to generate the temporary Spatial-Temporal information of a certain moment. Then, the Spatial-Temporal information of multiple single moments enters Temporal Pyramid Net (TPN) to generate the local Spatial-Temporal information of multiple time scales. Finally, bidirectional recurrent neutral network is used to act on the Spatial-Temporal information of all parts so as to acquire the context information spanning the length of the entire video, which endows the network with video context information extraction capability. Through the experiments on the three video action recognitio common experimental data sets UCF101, UCF11, UCFSports, the Spatial-Temporal information deep fusion network proposed in this paper has a high correct recognition rate in the task of video action recognitio. [ABSTRACT FROM AUTHOR]

Temporal knowledge base exists on various fields. Take medical medicine field as example, diabetes is a typical chronic disease which evolves slowly. This paper starts from actual EMR data of hospitals by combination of experience and knowledge of clinical doctors. Link prediction on clinical knowledge base such as diabetic complication requires the analysis on temporal characteristic of temporal knowledge base, which is a great challenge for traditional link prediction models. This paper proposes temporal knowledge graph link prediction model based on deep learning. This model selects the TransR transformation model suitable for big data and makes entity projection in relation space containing different semantic meanings, so as to vector the entities and complex semantic relations in graph. Then it adopts LSTM recursive neural network and adds the top-bottom relational information of the graph for sequential learning. Finally it constantly carries out deep learning through incremental calculation and LSTM recursive network to improve the accuracy of prediction. The incremental LSTM model highlights the hidden semantic and clinical temporal information and effectively utilizes sequential learning to mining forward-backward dependent information. It compensates the deficiency of lower prediction accuracy on timely knowledge graph caused by the traditional link prediction models. Finally, it is proved that the new model has better performance over temporal knowledge graph link prediction. [ABSTRACT FROM AUTHOR]

Learning based credit prediction has attracted great interest from academia and industry. Different institutions hold a certain amount of credit data with limited users to build model. An institution has the requirement to obtain data from other institutions for improving model performance. However, due to the privacy protection and subject to legal restrictions, they encounter difficulties in data exchange. This affects the performance of credit prediction. In order to solve the above problem, this paper proposes a federated learning based semi-supervised credit prediction approach enhanced by multi-layer label mean, which can aggregate parameters of each institution via joint training while protecting the data privacy of each institution. Moreover, in actual production and life, there are usually more unlabeled credit data than labeled ones, and the distribution of their feature space presents multiple data-dense divisions. To deal with these, local meanNet model is proposed with a multi-layer label mean based semi-supervised deep learning network. In addition, this paper introduces a cost-sensitive loss function in the supervised part of the local mean model. Conducted on two public credit datasets, experimental results show that our proposed federated learning based approach has achieved promising credit prediction performance in terms of Accuracy and F1 measures. At the same time, the framework design mode that splits data aggregation and keys uniformly can improve the security of data privacy and enhance the flexibility of model training. [ABSTRACT FROM AUTHOR]

In the deep learning-based video action recognitio, the function of the neural network is to acquire spatial information, motion information, and the associated information of the above two kinds of information over an uneven time span. This paper puts forward a network extracting video sequence semantic information based on deep integration of local Spatial-Temporal information. The network uses 2D Convolutional Neural Network (2DCNN) and Multi Spatial-Temporal scale 3D Convolutional Neural Network (MST_3DCNN) respectively to extract spatial information and motion information. Spatial information and motion information of the same time quantum receive 3D convolutional integration to generate the temporary Spatial-Temporal information of a certain moment. Then, the Spatial-Temporal information of multiple single moments enters Temporal Pyramid Net (TPN) to generate the local Spatial-Temporal information of multiple time scales. Finally, bidirectional recurrent neutral network is used to act on the Spatial-Temporal information of all parts so as to acquire the context information spanning the length of the entire video, which endows the network with video context information extraction capability. Through the experiments on the three video action recognitio common experimental data sets UCF101, UCF11, UCFSports, the Spatial-Temporal information deep fusion network proposed in this paper has a high correct recognition rate in the task of video action recognitio. [ABSTRACT FROM AUTHOR]

Temporal knowledge base exists on various fields. Take medical medicine field as example, diabetes is a typical chronic disease which evolves slowly. This paper starts from actual EMR data of hospitals by combination of experience and knowledge of clinical doctors. Link prediction on clinical knowledge base such as diabetic complication requires the analysis on temporal characteristic of temporal knowledge base, which is a great challenge for traditional link prediction models. This paper proposes temporal knowledge graph link prediction model based on deep learning. This model selects the TransR transformation model suitable for big data and makes entity projection in relation space containing different semantic meanings, so as to vector the entities and complex semantic relations in graph. Then it adopts LSTM recursive neural network and adds the top-bottom relational information of the graph for sequential learning. Finally it constantly carries out deep learning through incremental calculation and LSTM recursive network to improve the accuracy of prediction. The incremental LSTM model highlights the hidden semantic and clinical temporal information and effectively utilizes sequential learning to mining forward-backward dependent information. It compensates the deficiency of lower prediction accuracy on timely knowledge graph caused by the traditional link prediction models. Finally, it is proved that the new model has better performance over temporal knowledge graph link prediction. [ABSTRACT FROM AUTHOR]

To enable a quick and accurate access of targeted scientific and technological literature from massive stocks, here a deep content-based collaborative filtering method, namely DeepCCF, for personalized scientific and technological literature resources recommendation was proposed. By combining content-based filtering (CBF) and neural network-based collaborative filtering (NCF), the approach transforms the problem of scientific and technological literature recommendation into a binary classification task. Firstly, the word2vec is used to train the words embedding of the papers' titles and abstracts. Secondly, an academic literature topic model is built using term frequency–inverse document frequency (TF-IDF) and word embedding. Thirdly, the search and view history and published papers of researchers are utilized to construct the model that portrays the interests of researchers. Deep neural networks (DNNs) are then used to learn the nonlinear and complicated high-order interaction features between users and papers, and the top k recommendation list is generated by predicting the outputs of the model. The experimental results show that our proposed method can quickly and accurately capture the latent relations between the interests of researchers and the topics of paper, and be able to acquire the researchers' preferences effectively as well. The proposed method has tremendous implications in personalized academic paper recommendation, to propel technological progress. [ABSTRACT FROM AUTHOR]

Target detection is the basis for automatic target recognition system of infrared imaging guidance to complete subsequent tasks such as recognition and tracking. Existing systems have not the autonomous learning ability of target feature, and it will be powerless once the task environment exceeds the pre-planned condition. The single-stage target detection based on deep learning has the ability of autonomous learning and high computational efficiency, which is an effective way to solve the problem of infrared imaging guidance target detection in complex environment. SSD (Single Shot MultiBox Detector) is a classical single-stage detection model, however, the convolution layer with strong semantic information in SSD has low resolution, which is not conducive to small target detection. In addition, the location loss of SSD does not consider the impact of target scale change. Therefore, this paper puts forward two improvement ideas in view of SSD: (1) Starting from FPN (Feature Pyramid Network), feature channel's importance is distinguished through efficient channel attention mechanism, the contribution of each feature layer to the fusion output is described based on the learnable weight, and the feature weighted fusion of bidirectional multi-scale is realized between the feature layer which has low resolution and strong semantics and the feature layer which has high resolution and weak semantics. (2) Starting from IoU (Intersection over Union) and considering non overlapping parts and geometric relationship between the predicted box and the ground-truth box, the location loss of SSD that remains invariable to the target scale change is constructed to improve the sensitivity of the detection model to the locating error of small target. The experimental results show that, for 300 × 300 input, the presented method achieves 84.7% mAP (mean Average Precision) on VOC2007 test and for 512 × 512 input, it reaches 86.6%. On the self-built infrared aircraft data set, the proposed method achieves 81.1% mAP and can detect more small targets. Without affecting detection speed, the presented method on experimental results outperforms some comparable state-of-the-art models such as YOLOv3 (You Only Look Once), DSSD (Deconvolutional Single Shot Multibox Detector), RSSD (Rainbow Single Shot Multibox Detector) and FSSD (Fusion Single Shot Multibox Detector). [ABSTRACT FROM AUTHOR]

Traditional machine learning-based pest classification methods are a tedious and time-consuming process A method of multi-class pest detection based on deep learning and convolutional neural networks could be the solution. It automatically extracts the complex features of different pests from the crop pest images. In this paper, various significant deep learning-based object detection models like SSD, EfficientDet, Faster R-CNN, and CenterNet are implemented based on the Tensorflow Object Detection framework. Several significant networks like MobileNet_V2, ResNet101_V1, Inception_ResNet_V2, EfficientNet, and HourGlass104 are employed as backbone networks for these models to extract the different features of the pests. Object detection models are capable of identifying and locating pests in crops. Initially, these models are pre-trained with the COCO dataset and later be fine-tuned to the target pest dataset of 20 different pest classes. After conducting experiments on these models using the pest dataset, we demonstrate that Faster R-CNN_ResNet101_V1 outperformed every other model and achieved mAP of 74.77%. Additionally, it is developed as a lightweight model, whose size is ∼9 MB, and can detect pest objects in 130 milliseconds per image, allowing it to be used on resources-constrained devices commonly used by farmers. [ABSTRACT FROM AUTHOR]

Rain streaks severely affect the perception of the content and structure of an image so high-performance deraining algorithms are needed in order to eliminate the effects of various rain streaks for high-level computer vision tasks. Although much progress has been made with existing deraining methods, the task of single image deraining remains challenging. In this paper, we first point out that existing Transformers lack sufficient ability to capture channel attention which restricted the ability of models in deraining. To improve the performance of deraining model, we propose a dual branch deraining network based on Transformer. One branch uses dense connections to connect Transformer modules which embed the attention of a composite channel. This branch captures channel attention more finely to learn the representation of rain streaks features. The other branch first obtains features at different scales by gradually expanding the receptive field, then uses these features to obtain attention for regional features, and finally uses the attention to guide the model to focus on areas of high rain streaks density and large scales. By fusing these two branches, the model is able to capture channel attention more finely and to focus on regions of high rain streaks density and large scales. The extensive experimental results on synthetic and real datasets demonstrate that the proposed method outperforms most advanced deraining methods. [ABSTRACT FROM AUTHOR]

Power quality disturbance (PQD) defines the presence of inconsistencies that occur in the usual wave shapes of voltage and current signals. Power quality is considered the main challenge for power industry with the increase in dynamic load and highly subtle electronic devices. Besides, the islanding events, particularly unintended islanding, grasp significant challenges and it needs to be identified at the early stage. Islanding is an anomalousstate in the power system, where the distributed generators (DGs) are placed on supplying electrical energy to the local load even after the shortage of the major grid. Therefore, it is essential to identify and differentiate the PQ events and islanding events in ensuring pollution-free power, equipment, and labor safety. With this motivation, this paper presents an automated optimal deep learning based islanding detection (AODL-ID) technique. The proposed AODL-ID technique involves three major stages namely decomposition, classification, and hyperparameter tuning. Firstly, an empirical mode decomposition (EMD) approach is utilized to decompose the basic signals from the polluted signals. In addition, bidirectional gated recurrent neural network (BiGRNN) technique is employed for the classification of islanding and non-islanding PQ events in the wind energy penetrated DG systems by means of features (Voltage and current (RMS, half-cycle, peak and fundamental) Frequency. Power Factor / Cos Phi. Power and energy (active, reactive, harmonic, apparent)). Since the hyperparameters play a significant role in overall classification performance, the hyperparameter tuning of the BiGRNN model takes place using chaotic crow search algorithm (CCSA). To examine the enhanced classification outcome of the AODL-ID technique, a set of experimental analyses is carried out and the outcomes are investigated interms of various evaluation metrics. The simulation outcomes highlighted the supremacy of the AODL-ID technique over the compared techniques. [ABSTRACT FROM AUTHOR]

Aiming at the inconsistency of manual detection of mobile phone screen defects, the image feature extraction of traditional machine learning is often set based on experience, resulting in unsatisfactory detection results. Therefore, a mobile phone screen defect detection model (Ghostbackbone) which is proposed by this paper based on YOLOv5 s and Ghostbottleneck. The bottleneck of Ghostbackbone mainly uses and improves the Ghostbottleneck of GhostNet. The attention module of Ghostbackbone uses Coordinated Attention and Depthwise Separable Convolution for parameter reduction. Finally, Ghostbackbone uses YOLOv5 as the object detector to train the mobile phone screen defect dataset. The experimental results show that the parameter quantity of Ghostbackbone is 24% of that of YOLOv5 s, the average time of detecting a single picture is only 2% lower than that of YOLOv5 s, and the mAP0.5 : 0.95 is 2% higher than that of MobilenetV3 s. [ABSTRACT FROM AUTHOR]

In this information age, there is exponential growth in visual content and video captioning can address many real-life applications. Automatic generation of video captions can be beneficial to comprehend a video in a short time, assist in faster information retrieval, video analysis, indexing, report generation, etc. Captioning of industrial videos is of importance to get a visual and textual summary of the work ongoing in the industry. The generated captioned summary of the video can assist in remote monitoring of industries and these captions can be utilized for video question-answering, video segment extraction, productivity analysis, etc. Due to the presence of diverse events processing of industrial videos are more challenging compared to other domains. In this paper, we address the real-life application of generating the descriptions for the videos of a labor-intensive industry. We propose a keyframe-based approach for the generation of video captions. The framework produces a video summary by extraction of keyframes, thereby reducing the video captioning task to image captioning. These keyframes are passed to the image captioning model for description generation. Utilizing these individual frame captions, multi-caption descriptions of a video are generated with a unique start and end time of each caption. For image captioning, a merge encoder-decoder model with a stacked decoder for caption generation is used. We have performed experimentation on a dataset specifically created for the small-scale industry. We have also shown that data augmentation on the small dataset can greatly benefit the generation of remarkably good video descriptions. Results of extensive experimentation performed by utilizing different image encoders, language encoders, and decoders in the merge encoder-decoder model are reported. Apart from presenting the results on domain-specific data, results on domain-independent datasets are also presented to show the applicability of the technique in general. Performance comparison with existing datasets - OVSD and Flickr8k and Flickr30k are reported to demonstrate the scalability of our method. [ABSTRACT FROM AUTHOR]

The location attention mechanism has been widely applied in deep neural networks. However, as the mechanism entails heavy computing workload, significant memories consumed for weights storage, and shows poor parallelism in some calculations, it is hard to achieve high efficiency deployment. In this paper, the field-programmable gate array (FPGA) is employed to implement the location attention mechanism in hardware, and a novel fusion approach is proposed to connect the convolutional layer with the fully connected layer, which not only improves the parallelism of both the algorithm and the hardware pipeline, but also reduces the computation cost for such operations as multiplication and addition. Meanwhile, the shared computing architecture is used to reduce the demand for hardware resources. Parallel computing arrays are utilized to time-multiplex a single computing array, which can speed up the pipeline parallel computing of the attention mechanism. Experimental results show that for the location attention mechanism, the FPGA's inference speed is 0.010 ms, which is around a quarter of the speed achieved by running it with GPU, and its power consumption is 1.73 W, which is about 2.89% of the power consumed by running it with CPU. Compared with other FPGA implementation methods of attention mechanism, it has less hardware resource consumption and less inference time. When applied to speech recognition tasks, the trained attention model is symmetrically quantized and deployed on the FPGA. The result shows that the word error rate is only 0.79% higher than that before quantization, which proves the effectiveness and correctness of the hardware circuit. [ABSTRACT FROM AUTHOR]