Search

Employing deep learning techniques for the semantic segmentation of remote sensing images has emerged as a prevalent approach for acquiring information about water bodies. Yet, current models frequently fall short in accurately extracting water bodies from high-resolution remote sensing images, as these images often present intricate details of terrestrial objects and complex backgrounds. Vegetation, shadows, and other objects close to water boundaries have increased similarity to water bodies. Moreover, water bodies in high-resolution images have different boundary complexities, shapes, and sizes. This situation makes it somewhat challenging to accurately distinguish water bodies in high-resolution images. To overcome these difficulties, this paper presents a novel network model named EU-Net, specifically designed to extract water bodies from high-resolution remote sensing images. The proposed EU-Net model, with U-net as the backbone network, incorporates improved residual connections and attention mechanisms, and designs multi-scale dilated convolution and multi-scale feature fusion modules to enhance water body extraction performance in complex scenarios. Specifically, in the proposed model, improved residual connections are introduced to enable the learning of more complex features; the attention mechanism is employed to improve the model's discriminative ability by focusing on important channels and spatial areas. The implemented multi-scale dilated convolution technique enhances the model's receptive field while maintaining the same number of parameters. The designed multi-scale feature fusion module is capable of processing both small-scale details and large-scale structures in images, while simultaneously modeling the spatial context relationships of features at different scales. Experimental results validate the superior performance of EU-Net in accurately identifying water bodies from high-resolution remote sensing images, outperforming current models in terms of water extraction accuracy. [ABSTRACT FROM AUTHOR]

Abstract Employing deep learning techniques for the semantic segmentation of remote sensing images has emerged as a prevalent approach for acquiring information about water bodies. Yet, current models frequently fall short in accurately extracting water bodies from high-resolution remote sensing images, as these images often present intricate details of terrestrial objects and complex backgrounds. Vegetation, shadows, and other objects close to water boundaries have increased similarity to water bodies. Moreover, water bodies in high-resolution images have different boundary complexities, shapes, and sizes. This situation makes it somewhat challenging to accurately distinguish water bodies in high-resolution images. To overcome these difficulties, this paper presents a novel network model named EU-Net, specifically designed to extract water bodies from high-resolution remote sensing images. The proposed EU-Net model, with U-net as the backbone network, incorporates improved residual connections and attention mechanisms, and designs multi-scale dilated convolution and multi-scale feature fusion modules to enhance water body extraction performance in complex scenarios. Specifically, in the proposed model, improved residual connections are introduced to enable the learning of more complex features; the attention mechanism is employed to improve the model's discriminative ability by focusing on important channels and spatial areas. The implemented multi-scale dilated convolution technique enhances the model's receptive field while maintaining the same number of parameters. The designed multi-scale feature fusion module is capable of processing both small-scale details and large-scale structures in images, while simultaneously modeling the spatial context relationships of features at different scales. Experimental results validate the superior performance of EU-Net in accurately identifying water bodies from high-resolution remote sensing images, outperforming current models in terms of water extraction accuracy.

In this work, we propose an optimization approach for dynamic keystroke pattern recognition by leveraging a hybrid deep learning technique and multiple soft biometric factors. Our methodology begins with the introduction of a novel algorithm called dynamic drone squadron optimization (DDSO) to optimize the selection of optimal features from a pool of multiple keystroke features. We then present an enhanced version of the improved sperm swarm optimization (ISSO) algorithm, which effectively combines the optimal weight features derived from multiple biometric responses. Furthermore, we introduce the multi-stage recurrent neural network (MS-RNN) classifier to accurately recognize and classify keystroke patterns. The performance of our proposed ISSO+MS-RNN technique is evaluated using the benchmark KBOC dataset to validate its effectiveness. Comparative analysis is conducted against existing state-of-the-art techniques, employing various evaluation measures, to demonstrate the superior performance of proposed approach. [ABSTRACT FROM AUTHOR]

The exploration of drug combinations presents an opportunity to amplify therapeutic effectiveness while alleviating undesirable side effects. Nevertheless, the extensive array of potential combinations poses challenges in terms of cost and time constraints for experimental screening. Thus, it is crucial to narrow down the search space. Deep learning approaches have gained widespread popularity in predicting synergistic drug combinations tailored for specific cell lines in vitro settings. In the present study, we introduce a novel method termed GTextSyn, which utilizes the integration of gene expression data and chemical structure information for the prediction of synergistic effects in drug combinations. GTextSyn employs a sentence classification model within the domain of Natural Language Processing (NLP), wherein drugs and cell lines are regarded as entities possessing biochemical relevance. Meanwhile, combinations of drug pairs and cell lines are construed as sentences with biochemical relational significance. To assess the efficacy of GTextSyn, we conduct a comparative analysis with alternative deep learning approaches using a standard benchmark dataset. The results from a five-fold cross-validation demonstrate a 49.5% reduction in Mean Square Error (MSE) achieved by GTextSyn, surpassing the performance of the next best method in the regression task. Furthermore, we conduct a comprehensive literature survey on the predicted novel drug combinations and find substantial support from prior experimental studies for many of the combinations identified by GTextSyn. [ABSTRACT FROM AUTHOR]

Wireless sensor networks (WMSNs) are becoming increasingly popular in many fields, from academia to transportation, environmental monitoring, wildlife preservation, and military espionage. Therefore, examining potential threats, power consumption, vulnerability recognition, and systemic vulnerability characteristics is essential to develop a reliable information security approach for WSNs. As a result, it is becoming increasingly crucial for the technical community to conduct intrusion recognition method evaluations. Since this is the case, using deep learning techniques in creating intrusion identification and mitigation systems for wireless multimedia sensor networks is essential. This article examines how well different machine learning and deep learning algorithms perform in attack identification systems. Testing the efficacy of different methods on the WMSN-DS database through experimentation is essential. In this work, we combine the power of a Convolutional Neural Network classifier with a Random Forest. To accomplish this, a Convolutional Neural Network with a Random Forest Classifier is used. The intrusion detection system (IDS) is a crucial technique proposed in this study for WMSN. To address this issue, the current study proposal uses deep Learning with a Random Forest classifier to detect and prevent attacks and to promote efficient forwarding in WMSNs. Multiple WMSN assaults have been investigated, and the results of these investigations have been critically evaluated. [ABSTRACT FROM AUTHOR]

The recent advancements in artificial intelligence enhance the detection and classification of lung nodules via computed tomography scans, addressing the critical need for early diagnosis of lung cancer. The lung cancer when identified at the earlier stages, the chance of survival is higher. The methodology encompasses a modern deep-learning approach applied to a private dataset obtained from the Barnard Institute of Radiology at Madras Medical College, Chennai, which has been granted ethical approval. The results from applying the proposed Convolutional Neural Network model are promising, with an accuracy of 99.3% in malignancy detection, signifying a notable advancement in the precise diagnosis of lung cancer through non-invasive imaging techniques. Beyond academia, the findings of this study have significant implications for real-world healthcare settings. By providing a reliable and automated solution for lung nodule detection, this research contributes to early diagnosis and personalized treatment strategies for lung cancer patients. The value of the present work lies in its potential to reduce morbidity through the early detection of lung cancer, thus contributing to both clinical practice and the ongoing development of AI applications in healthcare. Our research may serve as a model for further studies in digital health care at Madras Medical College, aiming to improve patient outcomes through technology-driven diagnostics.

In many application domains, remotely sensed (RS) data are essential for disaster monitoring, climate forecasting, and remote surveillance. It perceives and gathers environmental, sociological, and transitional data on its own, from a distance, in increasing amounts of spatial and temporal details. While urbanization has been continued to be a worldwide phenomenon, the urban management and planning are needed to satisfy the needs of the Earth system. The operations associated with land management, urban planning, and transportation on Earth have been connected to digital maps and geographic databases. Knowledge of land cover is also significant for managing, conserving, and formulating environmental and urban development policies. Therefore, highly accurate Land Use Land Cover (LULC) assessment and monitoring are of the utmost significance for empirical studies and organizational accomplishment. In this study, we leverage remotely sensed data to formulate a sophisticated deep learning system for comprehensive Earth observation. The core of our approach involves the development of a three-layered Convolution-Convolution-Maxpooling Convolutional Neural Network (CCMCNN) model, preceded by a meticulous pre-processing stage. In the pre-processing stage, images extracted from the NWPU-RESISC dataset undergo resizing, division, and augmentation procedures before being fed into the three-layered CCMCNN architecture for the purpose of classifying them into five different LULC categories. The effectiveness of our proposed model undergoes thorough validation through a comparative examination against various pre-trained deep learning models outlined in existing literature. Presenting performance metrics, our model attains a remarkable accuracy rate of 98.4%. This advancement holds significant promise in the realm of Earth surveillance, offering a valuable tool for enhancing situational awareness and environmental monitoring.

Privacy and trust are significant issues in intelligent transportation systems (ITS). Data security is critical in ITS systems since sensitive user data is communicated to another user over the internet through wireless devices and routes such as radio channels, optical fiber, and blockchain technology. The Internet of Things (IoT) is a network of connected, interconnected gadgets. Privacy issues occasionally arise due to the amount of data generated.However, they have been primarily addressed by blockchain and smart contract technology. While there are still security issues with smart contracts, primarily due to the complexity of writing the code, there are still many challenges to consider when designing blockchain designs for the IoT environment. This study uses traditional blockchain technology with the “You Only Look Once” (YOLO) object detection method to accurately locate and identify license plates. While YOLO and blockchain technologies used for intelligent vehicle license plate recognition are promising, they have received limited research attention. Real-time object identification and recognition would be possible by combining a cutting-edge object detection technique with a regional convolutional neural network (RCNN) built with the tensor flow core open source libraries. This method works reasonably well for identifying any license plate. The Automatic License Plate Recognition (ALPR) approach delivered outstanding results in various datasets. First, with a recognition rate of 96.2%, our system (UFPR-ALPR) surpassed the previously used technology, consisting of 4500 frames and around 150 films. Second, a deep learning algorithm was trained to recognize images of license plate numbers using the UFPR-ALPR dataset. Third, the license plate’s characters were complicated for standard methods to identify because of the shifting lighting correctly. The proposed model, however, produced beneficial outcomes. [ABSTRACT FROM AUTHOR]

Deep learning (DL), a branch of machine learning and artificial intelligence, is nowadays considered as a core technology. Due to its ability to learn from data, DL technology originated from artificial neural networks and has become a hot topic in the context of computing, it is widely applied in various application areas. However, building an appropriate DL model is a challenging task, due to the dynamic nature and variations in real-world problems and data. The aim of this work was to develope a new method for appropriate DL model selection using complex spherical fuzzy rough sets (CSFRSs). The connectivity of two or more complex spherical fuzzy rough numbers can be defined by using the Hamacher t-norm and t-conorm. Using the Hamacher operational laws with operational parameters provides exceptional flexibility in dealing with uncertainty in data. We define a series of Hamacher averaging and geometric aggregation operators for CSFRSs, as well as their fundamental properties, based on the Hamacher t-norm and t-conorm. Further we have developed the proposed aggregation operators and provide here a group decision-making approach for solving decision making problems. Finally, a comparative analysis with existing methods is given to demonstrate the peculiarity of our proposed method. [ABSTRACT FROM AUTHOR]

Deep learning techniques have revolutionized the field of artificial intelligence by enabling accurate predictions of complex natural scenarios. This paper proposes a novel convolutional neural network (CNN) model that involves deep learning technologies, such as the bottleneck residual block, layer normalization, and dropout layer, to predict wave overtopping at coastal structures under a wide range of conditions. To optimize the performance of the CNN model, the hyperparameter tuning process via Bayesian optimization is used. The results of validation demonstrate that the proposed CNN model is highly accurate in estimating wave overtopping discharge from hydraulic and structural parameters. The testing accuracy of the overtopping predictions using a prototype dataset shows that the proposed CNN model outperforms those existing machine learning models. An example application of the CNN model is presented for predicting prototype overtopping considering various crest freeboards of coastal structures. [ABSTRACT FROM AUTHOR]

Deep learning (DL), a branch of machine learning and artificial intelligence, is nowadays considered as a core technology. Due to its ability to learn from data, DL technology originated from artificial neural networks and has become a hot topic in the context of computing, it is widely applied in various application areas. However, building an appropriate DL model is a challenging task, due to the dynamic nature and variations in real-world problems and data. The aim of this work was to develope a new method for appropriate DL model selection using complex spherical fuzzy rough sets (CSFRSs). The connectivity of two or more complex spherical fuzzy rough numbers can be defined by using the Hamacher t-norm and t-conorm. Using the Hamacher operational laws with operational parameters provides exceptional flexibility in dealing with uncertainty in data. We define a series of Hamacher averaging and geometric aggregation operators for CSFRSs, as well as their fundamental properties, based on the Hamacher t-norm and t-conorm. Further we have developed the proposed aggregation operators and provide here a group decision-making approach for solving decision making problems. Finally, a comparative analysis with existing methods is given to demonstrate the peculiarity of our proposed method.

Feces is one quick way to determine the health status of the birds and farmers rely on years of experience as well as professionals to identify and diagnose poultry diseases. Most often, farmers lose their flocks as a result of delayed diagnosis or a lack of trustworthy experts. Prevalent diseases affecting poultry birds may be quickly noticed from image of poultry bird's droppings using artificial intelligence based on computer vision and image analysis. This paper provides description of a dataset of both healthy and unhealthy poultry fecal imagery captured from selected poultry farms in south-west of Nigeria using smartphone camera. The dataset was collected at different times of the day to account for variability in light intensity and can be applied in machine learning models development for abnormality detection in poultry farms. The dataset collected is 19,155 images; however, after preprocessing which encompasses cleaning, segmentation and removal of duplicates, the data strength is 14,618 labeled images. Each image is 100 by 100 pixels size in jpeg format. Additionally, computer vision applications like picture segmentation, object detection, and classification can be supported by the dataset. This dataset's creation is intended to aid in the creation of comprehensive tools that will aid farmers and agricultural extension agents in managing poultry farms in an effort to minimize loss and, as a result, optimize profit as well as the sustainability of protein sources.

Vegetables are the staple food in our diets. Vegetable prices are difficult to forecast because they are influenced by a variety of factors, including weather, demand and supply chain, Government policies, etc. and exhibit volatile fluctuations. Marketing of vegetables is complex, especially because of their perishability, seasonality and bulkiness. An accurate and timely forecast of vegetables is essential to help its stakeholders. Previous studies observed that traditional statistical models are unable to capture the complex behaviour of vegetable markets. In this study, a comparative assessment has been carried out among the traditional time-series model, machine learning and deep learning techniques in order to find the best-suited model. For empirical illustration, cauliflower markets have been chosen as it is one of India's most important and popular winter. In order to identify the complexity in the price of cauliflower, the machine learning technique, i.e. artificial neural network and deep learning technique, i.e. long short-term memory model have been implemented. In addition, the traditional stochastic time-series model, i.e. autoregressive integrated moving average model, was used to compare the prediction accuracy of the above models. To this end, the moving window forecast approach was also implemented to evaluate the sensitivity of these models with respect to forecast length. It can be concluded that the deep learning model outperforms the traditional time-series model and the machine learning technique for both short- and long-term forecasting. [ABSTRACT FROM AUTHOR]

At an early point, the diagnosis of pancreatic cancer is mediocre, since the radiologist is skill deficient. Serious threats have been posed due to the above reasons, hence became mandatory for the need of skilled technicians. However, it also became a time-consuming process. Hence the need for automated diagnosis became mandatory. In order to identify the tumor accurately, this research proposes a novel Convolution Neural Network (CNN) based superior image classification technique. The proposed deep learning classification strategy has a precision of 97.7%, allowing for more effective usage of the automatically executed feature extraction technique to diagnose cancer cells. Comparative analysis with CNN-Grey Wolf Optimization (GWO) is carried based on varied testing and training outcomes. The suggested study is carried out at a rate of 90%-10%, 80%-20%, and 70%-30%, indicating the robustness of the proposed research work. Outcomes show that the suggested method is effective. GWO-CNN is reliable and accurate relative to other detection methods available in the literatures. [ABSTRACT FROM AUTHOR]

The coronavirus (COVID-19) is a lethal virus causing a rapidly infectious disease throughout the globe. Spreading awareness, taking preventive measures, imposing strict restrictions on public gatherings, wearing facial masks, and maintaining safe social distancing have become crucial factors in keeping the virus at bay. Even though the world has spent a whole year preventing and curing the disease caused by the COVID-19 virus, the statistics show that the virus can cause an outbreak at any time on a large scale if thorough preventive measures are not maintained accordingly. To fight the spread of this virus, technologically developed systems have become very useful. However, the implementation of an automatic, robust, continuous, and lightweight monitoring system that can be efficiently deployed on an embedded device still has not become prevalent in the mass community. This paper aims to develop an automatic system to simultaneously detect social distance and face mask violation in real-time that has been deployed in an embedded system. A modified version of a convolutional neural network, the ResNet50 model, has been utilized to identify masked faces in people. You Only Look Once (YOLOv3) approach is applied for object detection and the DeepSORT technique is used to measure the social distance. The efficiency of the proposed model is tested on real-time video sequences taken from a video streaming source from an embedded system, Jetson Nano edge computing device, and smartphones, Android and iOS applications. Empirical results show that the implemented model can efficiently detect facial masks and social distance violations with acceptable accuracy and precision scores. [ABSTRACT FROM AUTHOR]

Diagnosing a brain tumor takes a long time and relies heavily on the radiologist's abilities and experience. The amount of data that must be handled has increased dramatically as the number of patients has increased, making old procedures both costly and ineffective. Many researchers investigated a variety of algorithms for detecting and classifying brain tumors that were both accurate and fast. Deep Learning (DL) approaches have recently been popular in developing automated systems capable of accurately diagnosing or segmenting brain tumors in less time. DL enables a pre-trained Convolutional Neural Network (CNN) model for medical images, specifically for classifying brain cancers. The proposed Brain Tumor Classification Model based on CNN (BCM-CNN) is a CNN hyperparameters optimization using an adaptive dynamic sine-cosine fitness grey wolf optimizer (ADSCFGWO) algorithm. There is an optimization of hyperparameters followed by a training model built with Inception-ResnetV2. The model employs commonly used pre-trained models (Inception-ResnetV2) to improve brain tumor diagnosis, and its output is a binary 0 or 1 (0: Normal, 1: Tumor). There are primarily two types of hyperparameters: (i) hyperparameters that determine the underlying network structure; (ii) a hyperparameter that is responsible for training the network. The ADSCFGWO algorithm draws from both the sine cosine and grey wolf algorithms in an adaptable framework that uses both algorithms' strengths. The experimental results show that the BCM-CNN as a classifier achieved the best results due to the enhancement of the CNN's performance by the CNN optimization's hyperparameters. The BCM-CNN has achieved 99.98% accuracy with the BRaTS 2021 Task 1 dataset. [ABSTRACT FROM AUTHOR]

Deep learning techniques have revolutionized the field of artificial intelligence by enabling accurate predictions of complex natural scenarios. This paper proposes a novel convolutional neural network (CNN) model that involves deep learning technologies, such as the bottleneck residual block, layer normalization, and dropout layer, to predict wave overtopping at coastal structures under a wide range of conditions. To optimize the performance of the CNN model, the hyperparameter tuning process via Bayesian optimization is used. The results of validation demonstrate that the proposed CNN model is highly accurate in estimating wave overtopping discharge from hydraulic and structural parameters. The testing accuracy of the overtopping predictions using a prototype dataset shows that the proposed CNN model outperforms those existing machine learning models. An example application of the CNN model is presented for predicting prototype overtopping considering various crest freeboards of coastal structures.

Blood vessels are harmed in diabetic retinopathy (DR), a condition that impairs vision. Using modern healthcare research and technology, artificial intelligence and processing units are used to aid in the diagnosis of this syndrome and the study of diagnostic procedures. The correct assessment of DR severity requires the segmentation of lesions from fundus pictures. The manual grading method becomes highly difficult and time-consuming due to the wide range of the morphologies, number, and sizes of lesions. For image segmentation, traditional fuzzy clustering techniques have two major drawbacks. First, fuzzy memberships based clustering are more susceptible to outliers. Second, because of the lack of local spatial information, these techniques often result in oversegmentation of images. In order to address these issues, this research study proposes an outlier-based skimpy regularization fuzzy clustering technique (OSR-FCA) for image segmentation. Clustering methods that use fuzzy membership with sparseness can be improved by incorporating a Gaussian metric regularisation into the objective function. The proposed study used the symmetry information contained in the image data to conduct the image segmentation using the fuzzy clustering technique while avoiding over segmenting relevant data. This resulted in a reduced proportion of noisy data and better clustering results. The classification was carried out by a deep learning technique called convolutional neural network (CNN). Two publicly available datasets were used for the validation process by using different metrics. The experimental results showed that the proposed segmentation technique achieved 97.16% and classification technique achieved 97.26% of accuracy on the MESSIDOR dataset. [ABSTRACT FROM AUTHOR]

During its long service life, an arch dam affected by a combination of factors exhibits a typical time-varying characteristic in terms of its structure and material properties, and the deformation in the dam structure can directly and reliably reflect the health and service status of dams. Therefore, an accurate deformation prediction is an important part of dam safety monitoring. However, due to multiple factors, dam deformation data often tend to be highly volatile, and most existing deformation estimation techniques employ a single algorithm, which may not effectively capture the potential change process. A hybrid model for dam deformation prediction has been proposed to overcome this problem. First, dam deformation data are decomposed into three components by seasonal and trend decomposition using loess. Second, a convolutional neural network–gated recurrent unit (GRU) hybrid model, which optimizes hyperparameters using the sparrow search algorithm, is used to capture the nonlinear relationships that exist in each component. Finally, the final prediction result of dam deformation is the comprehensive output of multiple submodules. The deformation monitoring data (period: 2009–2019) of a parabolic variable-thickness double-curved arch dam located in China are considered as the survey target. The test results indicate that the proposed model is suitable for short-term and long-term prediction and outperforms other models in terms of higher robustness to abnormal sequences than other conventional models (R² differs by 5.50% and 7.87%, respectively, in short-term and long-term predictions for different measurement points, while other models differ by 9.78% to reach 15.71%, respectively). Among the models studied, the GRU shows better robustness to abnormal series than the LSTM with good prediction accuracy, fewer parameters, and a simpler structure. Hence, the GRU can be employed for dam deformation prediction in practical engineering. [ABSTRACT FROM AUTHOR]

The current study of kidney stones made by the National Health and Nutrition Examination Survey (NHANES) describes that over the last eighty years the number of stones present in the kidney is significantly enhanced by approximately 70% (2015). According to Malaysian Dialysis and information of Transplant Registry, there is a constant increase in the number of patients who are registering newly, and the increase from 2002 to 2010 is 2375, 5153 from 2011 to 2014, and from 2015 to 2016 it is 5201. A standardized data system in the United States (2008–2009). In the United States Renal Data System which captures, evaluates, and generates reports that are based on end-stage renal disease and diseases of the kidney. Some of the pathological changes of the kidney are the formation of stones, variations in the state of its position, and inflammation of the kidney. It is very essential to recognize the precise and correct position of stone in the kidney for interventional procedures. For further analysis, the ultrasound images are considered as a challenging task because of its significantly low intensity and additive sound. Therefore, to eliminate the additive noise present in the image, preprocessing of the US is performed. The preprocessed US image by using Plain Intensity Filter is fed to feature extraction technique extracted by deep learning-based advanced Symlets, Daubechies, and Biorthogonal wavelet sub-bands filters to extract kidney image and stones. These features are subjected to segmentation using a Novel Deep and Machine learning Reaction–Diffusion (RD) level set segmentation (LSS). It employs a two-step division process to recursively solve the equation of RD-LSS such as reaction and diffusion. The reaction is to minimize the numerical errors and the diffusion is employed to normalize the set function level for better stability. The RD-LSS also eliminates the anti-leakages on image boundary without requiring any expensive re-initialization. Wavelet sub-bands are used to obtain energy and mean levels and the average levels provide a sign of the absence or existence of stones. These levels of energy are accomplished through Machine learning algorithms with the help of Back Propagation, Multilayer Perceptron, and Support Vector Machine ANN (Verma et al. in Pattern Recognit Image Anal 27:574–580, 35). After successfully training by Machine learning algorithms, the proposed system can identify the single and multiple stones with an accuracy of 98.9% in terms of detection of stone in US images compared with the existing detection techniques that give 92.1% accuracy. [ABSTRACT FROM AUTHOR]

Purpose: Nephron-sparing surgery (NSS) is a mainstream treatment for localized renal tumors. Segmental renal artery clamping (SRAC) is commonly used in NSS. Automatic and precise segmentations of renal artery trees are required to improve the workflow of SRAC in NSS. In this study, we developed a tridimensional kidney perfusion (TKP) model based on deep learning technique to automatically demonstrate renal artery segmentation, and verified the precision and feasibility during laparoscopic partial nephrectomy (PN). Methods: The TKP model was established based on convolutional neural network (CNN), and the precision was validated in porcine models. From April 2018 to January 2020, TKP model was applied in laparoscopic PN in 131 patients with T1a tumors. Demographics, perioperative variables, and data from the TKP models were assessed. Indocyanine green (ICG) with near-infrared fluorescence (NIRF) imaging was applied after clamping and dice coefficient was used to evaluate the precision of the model. Results: The precision of the TKP model was validated in porcine models with the mean dice coefficient of 0.82. Laparoscopic PN was successfully performed in all cases with segmental renal artery clamping (SRAC) under TKP model's guidance. The mean operation time was 100.8 min; the median estimated blood loss was 110 ml. The ischemic regions recorded in NIRF imaging were highly consistent with the perfusion regions in the TKP models (mean dice coefficient = 0.81). Multivariate analysis revealed that the feeding lobar artery number was strongly correlated with tumor size and contact surface area; the supplying segmental arteries number correlated with tumor size. Conclusions: Using the CNN technique, the TKP model is developed to automatically present the renal artery trees and precisely delineate the perfusion regions of different segmental arteries. The guidance of the TKP model is feasible and effective in nephron-sparing surgery. [ABSTRACT FROM AUTHOR]

In the arena of image forensics, detecting manipulations in an image is extremely significant because of the use of images in different fields. Various detection techniques have been suggested in the literature that are based on digging out the features from images to unveil the traces left by manipulation operations. In this paper, a deep learning-based approach is proposed in which a residual network is used to learn deep, complex features from preprocessed images for classification into authentic and forged images. There is statistical symmetry in similar types of images and asymmetry in different types of images. The proposed scheme can highlight the statistical asymmetry between authentic and forged images. In the proposed scheme, firstly, an RGB image is analyzed for different JPEG compression levels. The obtained difference between the error levels is used to extract enhanced LBP code. Then, the scale- and direction-invariant LBP (SD-LBP) code is transformed into SD-LBP feature maps to feed to a deep residual network. Next, the concept of explainable artificial intelligence (XAI) is used to help provide explanations and interpret the output, thereby raising the credibility of the proposed approach. The unique feature selection approach employed is the kernel SHAP method, which is focused on the Shapley values. This technique is used to pinpoint the specific characteristics that are responsible for the aberrant behavior of the forged images dataset. Later, the deep learning-based model is trained and validated using these feature sets. A pre-activation version of ResNet-50 architecture is used that achieved an accuracy of 99.31%, 99.52%, 98.05%, and 99.10% on CASIA v1, CASIA v2, IMD 2020, and DVMM datasets, respectively. The capability of the pretrained residual network and rich textural features, which are scale- and direction-invariant, helps to expand the detection accuracy of the proposed approach. The results confirmed that the method either produced competitive results or outperformed existing methods. [ABSTRACT FROM AUTHOR]

Digital image forensics has become necessary as an emerging technology. Images can be adulterated effortlessly using image tools. The latest techniques are available to detect whether an image is adulterated by a particular operator. Most of the existing techniques are suitable for high resolution and manipulated images by a single operator. In a real scenario, multiple operators are applied to manipulate the image many times. In this paper, a robust moderate-sized convolutional neural network is proposed to identify manipulation operators and also the operator's sequence for two operators in particular. The proposed bottleneck approach is used to make the network deeper and reduce the computational cost. Only one pooling layer, called a global averaging pooling layer, is utilized to retain the maximum flow of information and to avoid the overfitting issue between the layers. The proposed network is also robust against low resolution and JPEG compressed images. Even though the detection of the operator is challenging due to the limited availability of statistical information in low resolution and JPEG compressed images, the proposed model can also detect an operator with different parameters and compression quality factors that are not considered in training. [ABSTRACT FROM AUTHOR]

Reports indicate that high-cost, insecurity, and difficulty in complex environments hinder the traditional urban road inundation monitoring approach. This work proposed an automatic monitoring method for experimental urban road inundation based on the YOLOv2 deep learning framework. The proposed method is an affordable, secure, with high accuracy rates in urban road inundation evaluation. The automatic detection of experimental urban road inundation was carried out under both dry and wet conditions on roads in the study area with a scale of a few m2. The validation average accuracy rate of the model was high with 90.1% inundation detection, while its training average accuracy rate was 96.1%. This indicated that the model has effective performance with high detection accuracy and recognition ability. Besides, the inundated water area of the experimental inundation region and the real road inundation region in the images was computed, showing that the relative errors of the measured area and the computed area were less than 20%. The results indicated that the proposed method can provide reliable inundation area evaluation. Therefore, our findings provide an effective guide in the management of urban floods and urban flood-warning, as well as systematic validation data for hydrologic and hydrodynamic models. HIGHLIGHTS First experimental urban road inundation automatic detection study using YOLOv2.; Proposed an inundation area computation method based on a deep learning technique.; Good performance on an experimental urban road inundation detection was tested.;

In view of the problems of high cost and low reliability in obtaining operation information such as flow rate and pressure of offshore natural gas production system, research on soft sensing is carried out, and a dynamic data‐driven model bank is established, in purpose of estimating single‐well flow rate and wellhead pressure, providing convenience tool for online monitoring and system safety analysis. Combining dynamic and steady‐state samples, introducing black‐box identification techniques including orthogonal least square regression and deep learning along with parameter correction techniques such as bi‐objective least square algorithm and transfer learning, a series of nonlinear auto‐regressive models with exogenous inputs (NARX) are built, consisting of black‐box and gray‐box polynomial NARX (Poly‐NARX) models as well as deep neural network NARX (DNN‐NARX) models, approximately describing the dynamic performance of gas production well. Through realistic operation data, the simulation results of Poly‐NARX, DNN‐NARX, and multiple‐layer‐perception‐NARX models are compared. It is observed that gray‐box DNN‐NARX model shows the best performance with advantages of higher global applicability, better approximation ability, and stronger generalization ability. Proposed model bank is of high expansibility and engineering applicability for soft sensing problems in the petroleum industry, laying the ground work for building smart oil and gas field. [ABSTRACT FROM AUTHOR]

The moisture content of tobacco, as an important characteristic which should be kept at a desired level to maintain consistent product quality in drying process, is difficult to perform the direct measurement and anomaly detection due to its large delay in actual process. Therefore, an intelligent real-time detection method is an urgent and challenging task in ensuring the product quality. This paper proposes a time-domain raw data conversion method along with a novel deep learning architecture called multi-hierarchical convolutional neural network (MHCNN) for moisture prediction, in which the proposed architecture automatically learns multi-hierarchical features from transformed image-like data and simultaneously performs online prediction. Experiments are conducted on the real production data from the cigarette factory and the presented model performs well on overall testing dataset. Specifically, the MAE, RMSE and R2 of normal production batch can reach to 0.0131, 0.0244, and 0.9721 respectively, which are far superior to the estimation of experience and other alternatives. It demonstrates that the proposed online prediction strategy can simultaneously perform multi-hierarchical feature extraction and moisture online prediction with high precise to eliminate the detection delay for process optimization and control. [ABSTRACT FROM AUTHOR]

Abstract In view of the problems of high cost and low reliability in obtaining operation information such as flow rate and pressure of offshore natural gas production system, research on soft sensing is carried out, and a dynamic data‐driven model bank is established, in purpose of estimating single‐well flow rate and wellhead pressure, providing convenience tool for online monitoring and system safety analysis. Combining dynamic and steady‐state samples, introducing black‐box identification techniques including orthogonal least square regression and deep learning along with parameter correction techniques such as bi‐objective least square algorithm and transfer learning, a series of nonlinear auto‐regressive models with exogenous inputs (NARX) are built, consisting of black‐box and gray‐box polynomial NARX (Poly‐NARX) models as well as deep neural network NARX (DNN‐NARX) models, approximately describing the dynamic performance of gas production well. Through realistic operation data, the simulation results of Poly‐NARX, DNN‐NARX, and multiple‐layer‐perception‐NARX models are compared. It is observed that gray‐box DNN‐NARX model shows the best performance with advantages of higher global applicability, better approximation ability, and stronger generalization ability. Proposed model bank is of high expansibility and engineering applicability for soft sensing problems in the petroleum industry, laying the ground work for building smart oil and gas field.

[Display omitted] • A non-noble-metal based novel SERS substrate-porous carbon film/WO 3-x nanosheets was developed; • Detection limit of 10-7 M with the Raman enhancement factor of 106 was achieved on the optical substrate. • A deep leaning method was used to accurately analyze the Raman spectra of flavonoids of quercetin, 3-hydroxyflavone and flavone. • Quercetin can be accurately detected by the fully convolutional network, and lowest detectable concentration of 10-5 M was achieved. The Surface-enhanced Raman scattering (SERS) is an attractive optical detecting method with high sensitivity and detectivity, however challenges on large-area signal uniformity and complex spectra analysis methods always retards its wide application. Herein, a highly sensitive and uniform SERS detection strategy supported by porous carbon film/WO 3-x nanosheets (PorC/WO 3-x) based noble-metal-free SERS substrate and deep learning algorithm are reported. Experimentally, the PorC/WO 3-x substrate was prepared by high-temperature annealing the PorC/WO 3 films under the argon atmosphere. The defect density of the WO 3 was controlled by tuning the reducing reaction time during the annealing process. The SERS performance was evaluated by using R6G as the Raman reporter, it showed that the SERS intensity obtained on the substrate with the optimal annealing time of 3 h was about 8 times as high as that obtained on the PorC/WO 3 substrate without annealing treatment. And detection limit of 10-7 M and Raman enhancement factor of 106 could be achieved. Moreover, the above optimal SERS substrate was utilized to detect flavonoids of quercetin, 3-hydroxyflavone and flavone, and a deep learning algorithms was incorporated to identify the quercetin. It revealed that quercetin can be accurately detected within the above flavonoids, and lowest detectable concentration of 10-5 M can be achieved. [ABSTRACT FROM AUTHOR]

The fog computing model has emerged as a viable infrastructure for serving IoT applications in recent years. In the fog ecosystem, it is essential to manage resources for different workloads due to the high volume and rapid growth of requests. Therefore, a challenge faced in this area is dynamic and efficient resource auto-scaling because fog resources must be allocated to requests efficiently. More fog resources than needed leads to "Over-Provisioning", and fewer fog resources leads to the "Under-provisioning" issue. To this end, an effective deep learning-based resource auto-scaling mechanism has been proposed to manage the number of resources needed to handle dynamic workloads in a fog environment. The simulation results indicated that the proposed solution reduces cost, network usage, and delay violation and increases CPU utilization compared with existing resource auto-scaling mechanisms. [ABSTRACT FROM AUTHOR]

Diagnosing a brain tumor takes a long time and relies heavily on the radiologist’s abilities and experience. The amount of data that must be handled has increased dramatically as the number of patients has increased, making old procedures both costly and ineffective. Many researchers investigated a variety of algorithms for detecting and classifying brain tumors that were both accurate and fast. Deep Learning (DL) approaches have recently been popular in developing automated systems capable of accurately diagnosing or segmenting brain tumors in less time. DL enables a pre-trained Convolutional Neural Network (CNN) model for medical images, specifically for classifying brain cancers. The proposed Brain Tumor Classification Model based on CNN (BCM-CNN) is a CNN hyperparameters optimization using an adaptive dynamic sine-cosine fitness grey wolf optimizer (ADSCFGWO) algorithm. There is an optimization of hyperparameters followed by a training model built with Inception-ResnetV2. The model employs commonly used pre-trained models (Inception-ResnetV2) to improve brain tumor diagnosis, and its output is a binary 0 or 1 (0: Normal, 1: Tumor). There are primarily two types of hyperparameters: (i) hyperparameters that determine the underlying network structure; (ii) a hyperparameter that is responsible for training the network. The ADSCFGWO algorithm draws from both the sine cosine and grey wolf algorithms in an adaptable framework that uses both algorithms’ strengths. The experimental results show that the BCM-CNN as a classifier achieved the best results due to the enhancement of the CNN’s performance by the CNN optimization’s hyperparameters. The BCM-CNN has achieved 99.98% accuracy with the BRaTS 2021 Task 1 dataset.

During its long service life, an arch dam affected by a combination of factors exhibits a typical time-varying characteristic in terms of its structure and material properties, and the deformation in the dam structure can directly and reliably reflect the health and service status of dams. Therefore, an accurate deformation prediction is an important part of dam safety monitoring. However, due to multiple factors, dam deformation data often tend to be highly volatile, and most existing deformation estimation techniques employ a single algorithm, which may not effectively capture the potential change process. A hybrid model for dam deformation prediction has been proposed to overcome this problem. First, dam deformation data are decomposed into three components by seasonal and trend decomposition using loess. Second, a convolutional neural network–gated recurrent unit (GRU) hybrid model, which optimizes hyperparameters using the sparrow search algorithm, is used to capture the nonlinear relationships that exist in each component. Finally, the final prediction result of dam deformation is the comprehensive output of multiple submodules. The deformation monitoring data (period: 2009–2019) of a parabolic variable-thickness double-curved arch dam located in China are considered as the survey target. The test results indicate that the proposed model is suitable for short-term and long-term prediction and outperforms other models in terms of higher robustness to abnormal sequences than other conventional models (R² differs by 5.50% and 7.87%, respectively, in short-term and long-term predictions for different measurement points, while other models differ by 9.78% to reach 15.71%, respectively). Among the models studied, the GRU shows better robustness to abnormal series than the LSTM with good prediction accuracy, fewer parameters, and a simpler structure. Hence, the GRU can be employed for dam deformation prediction in practical engineering.

This work presents an experiment involving detailed fluvial flood propagation process. Comparing to the existing flood experiments which collect hydrodynamical data just at gauges, flood evolution process in river channel and flood plain is measured and temporal–spatial data are provided. In the experiment, three inflow patterns are considered to reflect the different severity of the floods. The flood propagation and inundation are captured by using an array of surveillance cameras. The images are pre‐processed by applying camera calibration method to correct the barrel distortion. A deep learning technique is introduced to automatically identify inundated area. The inundation process is therefore obtained by identifying image series. In addition to the spatial data, the water level evolutions at three gauges are also monitored to supply detailed hydrodynamic information at gauges. The repeatability of the experiments and reliability of the deep learning technique are verified. The experimental data including spatial and point hydrodynamic features for flood events can be used to systematically validate numerical model and calibrate parameters. [ABSTRACT FROM AUTHOR]

In this paper, a survey has been performed on deep learning techniques for supporting farmers in monitoring the crops and their yields. The study has been done on the specific models and frameworks employed for agriculture issues. Also, a study has been done on the data set used, and the overall performance achieved according to the metrics. The comparison listed on basis of few parameters and the findings indicate that deep learning provides high accuracy and is commonly used for image processing techniques. This survey will surely be beneficial for farmers and researchers to find out the results about trends and statistical information of crops, updating facilities and land availability which can in turn help in improving production. [ABSTRACT FROM AUTHOR]

Extensive use of multimedia services and Internet Data Center applications demand distributed deployment of these applications. It is implemented using edge computing with server clusters. To increase the availability of the services, applications are deployed redundantly in server clusters. In this situation, an efficient server allocation strategy is essential to improve execution fairness in server cluster. Categorizing the incoming traffic at server cluster is desired for the improvement of QoS. The traditional traffic classification models categorize the incoming traffic according to their applications' type. They are ineffective in selection of suitable server, as they do not consider the characteristics of the server. Hence this paper proposes a classifier to assist the dispatcher to distribute the requests to appropriate server in server cluster. The proposed deep learning classification model based on incoming traffic characteristics and server status is reinforced with extended labelling using correlation based approach. The experimental results of the proposed classifier have shown considerable performance enhancement in terms of classification measures and waiting time of the requests compared to existing machine learning models. [ABSTRACT FROM AUTHOR]

The numerical study of electrocardiology involves prohibitive computational costs because of its complex and nonlinear dynamics. In this paper, a lowdimensional model of the cardiac monodomain formulation has been developed by using the deep learning method. The restricted Boltzmann machine and deep autoencoder machine learning techniques have been used to approximate the cardiac tissue's full order dynamics. The proposed reduced order modeling begins with the development of the low-dimensional representations of the original system by implementing the neural networks from the numerical simulations of the full order monodomain system. Consequently, the reduced order representations have been utilized to construct the lower-dimensional model, and finally, it has been reconstructed back to the original system. Numerical results show that, the proposed deep learning MOR framework gained computational efficiency by a factor of 85 with acceptable accuracy. This paper compares the accuracy of the deep learning based model order reduction method with the two different techniques of model reduction: proper orthogonal decomposition (POD) and dynamic mode decomposition (DMD). The model order reduction deploying the deep learning method outperforms the POD and DMD concerning the modeling accuracy. [ABSTRACT FROM AUTHOR]

Abstract This work presents an experiment involving detailed fluvial flood propagation process. Comparing to the existing flood experiments which collect hydrodynamical data just at gauges, flood evolution process in river channel and flood plain is measured and temporal–spatial data are provided. In the experiment, three inflow patterns are considered to reflect the different severity of the floods. The flood propagation and inundation are captured by using an array of surveillance cameras. The images are pre‐processed by applying camera calibration method to correct the barrel distortion. A deep learning technique is introduced to automatically identify inundated area. The inundation process is therefore obtained by identifying image series. In addition to the spatial data, the water level evolutions at three gauges are also monitored to supply detailed hydrodynamic information at gauges. The repeatability of the experiments and reliability of the deep learning technique are verified. The experimental data including spatial and point hydrodynamic features for flood events can be used to systematically validate numerical model and calibrate parameters.

In the arena of image forensics, detecting manipulations in an image is extremely significant because of the use of images in different fields. Various detection techniques have been suggested in the literature that are based on digging out the features from images to unveil the traces left by manipulation operations. In this paper, a deep learning-based approach is proposed in which a residual network is used to learn deep, complex features from preprocessed images for classification into authentic and forged images. There is statistical symmetry in similar types of images and asymmetry in different types of images. The proposed scheme can highlight the statistical asymmetry between authentic and forged images. In the proposed scheme, firstly, an RGB image is analyzed for different JPEG compression levels. The obtained difference between the error levels is used to extract enhanced LBP code. Then, the scale- and direction-invariant LBP (SD-LBP) code is transformed into SD-LBP feature maps to feed to a deep residual network. Next, the concept of explainable artificial intelligence (XAI) is used to help provide explanations and interpret the output, thereby raising the credibility of the proposed approach. The unique feature selection approach employed is the kernel SHAP method, which is focused on the Shapley values. This technique is used to pinpoint the specific characteristics that are responsible for the aberrant behavior of the forged images dataset. Later, the deep learning-based model is trained and validated using these feature sets. A pre-activation version of ResNet-50 architecture is used that achieved an accuracy of 99.31%, 99.52%, 98.05%, and 99.10% on CASIA v1, CASIA v2, IMD 2020, and DVMM datasets, respectively. The capability of the pretrained residual network and rich textural features, which are scale- and direction-invariant, helps to expand the detection accuracy of the proposed approach. The results confirmed that the method either produced competitive results or outperformed existing methods.