Search

Micro-expression (ME) can reflect a person's authentic internal thoughts and thus have significant research value in many fields. With the continuous development of computer technology, ME analysis methods have gradually shifted from psychological-based to computer vision-based in recent years. ME spotting, a critical branch in ME analysis, has received more and more attention. Some review papers on ME have been published in recent years. However, most of them focused mainly on ME recognition and lacked a detailed study of ME spotting. Therefore, this paper attempts to conduct a comprehensive review of the latter. Specifically, we first summarize the research scope of ME spotting and introduce the current spontaneous ME datasets. Then, we review the ME spotting methods from two aspects: apex detection and interval detection. We further classify ME interval detection into handcrafted-based and deep learning-based. The characteristics and limitations of each technique are discussed in detail. After that, evaluation metrics and the experimental comparison of these methods followed. In addition, we discuss the challenges in ME spotting and outline the possible directions for future research. We hope this review article could assist researchers in better understanding ME spotting. [ABSTRACT FROM AUTHOR]

Electronic health records provide a vast amount of text health data written by physicians as patient clinical notes. The world health organization released the international classification of diseases version 10 (ICD-10) system to monitor and analyze clinical notes. ICD-10 is system physicians and other healthcare providers use to classify and code all diagnoses and symptom records in conjunction with hospital care. Therefore, the data can be easily stored, retrieved, and analyzed for decision-making. In order to address the problem, this paper introduces a system to classify the clinical notes to ICD-10 codes. This paper examines 7541 clinical notes collected from a health institute in Jordan and annotated by ICD-10's coders. In addition, the research uses another outsource dataset to augment the actual dataset. The research presented many approaches, such as the baseline and pipeline models. The Baseline model employed several methods like Word2vec embedding for representing the text. The model structure also involves long-short-term memory a convolutional neural network, and two fully-connected layers. The second Pipeline approach adopts the transformer model, such as Bidirectional Encoder Representations from Transformers (BERT), which is pre-trained on a similar health domain. The Pipeline model builds on two BERT models. The first model classifies the category codes representing the first three characters of ICD-10. The second BERT model uses the outputs from the general BERT model (first model) as input for the special BERT (second model) to classify the clinical notes into total codes of ICD-10. Moreover, Baseline and Pipeline models applied the Focal loss function to eliminate the imbalanced classes. However, The Pipeline model demonstrates a significant performance by evaluating it over the F1 score, recall, precision, and accuracy metric, which are 92.5%, 84.9%, 91.8%, and 84.97%, respectively. [ABSTRACT FROM AUTHOR]

A traditional approach for panoramic image generation consists of a random sample consensus (RANSAC) algorithm on a set of scale-invariant feature transform (SIFT) correspondences to generate a homography matrix between two images. Although producing adequate results for some type of images, hand-crafted SIFT features are not robust enough for highly varying natural images and the iterative RANSAC algorithm with its randomness does not always find the desired homography matrix. Recently, deep neural networks have been producing significant results in many challenging computer vision problems by learning features from large amounts of data. However, only very few recent works have been applied deep learning to panoramic image generation with the objective of finding feature correspondences and estimating homography matrix. Moreover, the absence of a proper dataset for the image stitching task hinders the standardization of models and comparison of their results. This paper attempts to generate panoramic images by extensively experimenting with various approaches using deep neural networks. The best proposed deep learning model achieved 7.31 and 1.07 pixels of the average absolute value loss for corner difference in X and Y directions, respectively. At the same time, qualitative results demonstrate superiority in comparison with the state-of-the-art SIFT+RANSAC algorithm. Specifically, in 72% of time the proposed framework either produced better results than SIFT+RANSAC or results of the proposed approach and SIFT+RANSAC were indistinguishable. Although SIFT + RANSAC produces better results in 28% of the time with respect to the loss function, our results are still visually comparable in many of these cases. Finally, a novel panoramic image generation dataset is introduced in this paper. [ABSTRACT FROM AUTHOR]

Pancreatic cancer is associated with higher mortality rates due to insufficient diagnosis techniques, often diagnosed at an advanced stage when effective treatment is no longer possible. Therefore, automated systems that can detect cancer early are crucial to improve diagnosis and treatment outcomes. In the medical field, several algorithms have been put into use. Valid and interpretable data are essential for effective diagnosis and therapy. There is much room for cutting-edge computer systems to develop. The main objective of this research is to predict pancreatic cancer early using deep learning and metaheuristic techniques. This research aims to create a deep learning and metaheuristic techniques-based system to predict pancreatic cancer early by analyzing medical imaging data, mainly CT scans, and identifying vital features and cancerous growths in the pancreas using Convolutional Neural Network (CNN) and YOLO model-based CNN (YCNN) models. Once diagnosed, the disease cannot be effectively treated, and its progression is unpredictable. That's why there's been a push in recent years to implement fully automated systems that can sense cancer at a prior stage and improve diagnosis and treatment. The paper aims to evaluate the effectiveness of the novel YCNN approach compared to other modern methods in predicting pancreatic cancer. To predict the vital features from the CT scan and the proportion of cancer feasts in the pancreas using the threshold parameters booked as markers. This paper employs a deep learning approach called a Convolutional Neural network (CNN) model to predict pancreatic cancer images. In addition, we use the YOLO model-based CNN (YCNN) to aid in the categorization process. Both biomarkers and CT image dataset is used for testing. The YCNN method was shown to perform well by a cent percent of accuracy compared to other modern techniques in a thorough review of comparative findings. [ABSTRACT FROM AUTHOR]

Haphazard urbanization has resulted in development of unplanned informal settlements, also called slums. Citizens residing in these dense settlements are generally deprived of essential sustainable drivers, i.e., green cover and open spaces (GOS). Such limitations can have an adverse impact on various aspects such as physiological, economical, health, and quality of life. Much research has been done on applying Machine Learning (ML) techniques for identifying slums at the city scale. However, no study explicitly addressed the problem of GOS extraction inside the informal settlements. This paper aims to fill this gap by training three modified Convolution Neural Network (CNN) models, i.e., VGG16-UNet, MobileNetV2-UNet, DeepLabV3 + on manually labeled high-resolution satellite imagery for the city of Mumbai. All three models performed excellently and achieved the state-of-the art overall accuracy of around 95%, for the considered classes. Out of the three models, VGG16-UNet performed the best and hence was applied to the whole city for developing green, open, and green and open indices. The indices were further used to establish a quantitative relation between various built-up typologies and GOS. It was found that informal settlements have significantly less green and open space than other built-up typologies. Stark GOS difference between slums and its neighborhood planned residential was observed, and hence, requires urgent attention of the city managers. The outcome presented in the paper can be used as input to various decision-making systems concerning better informal space planning to better plan the GOS, which eventually can lead to development of sustainable cities. [ABSTRACT FROM AUTHOR]

Coronavirus Disease 2019 (COVID-19), which is caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-COV-2), surprised the world in December 2019 and has threatened the lives of millions of people. Countries all over the world closed worship places and shops, prevented gatherings, and implemented curfews to stand against the spread of COVID-19. Deep Learning (DL) and Artificial Intelligence (AI) can have a great role in detecting and fighting this disease. Deep learning can be used to detect COVID-19 symptoms and signs from different imaging modalities, such as X-Ray, Computed Tomography (CT), and Ultrasound Images (US). This could help in identifying COVID-19 cases as a first step to curing them. In this paper, we reviewed the research studies conducted from January 2020 to September 2022 about deep learning models that were used in COVID-19 detection. This paper clarified the three most common imaging modalities (X-Ray, CT, and US) in addition to the DL approaches that are used in this detection and compared these approaches. This paper also provided the future directions of this field to fight COVID-19 disease. [ABSTRACT FROM AUTHOR]

As Earth observation technology has advanced, the volume of remote sensing big data has grown rapidly, offering significant obstacles to efficient and effective processing and analysis. A convolutional neural network refers to a neural network that covers convolutional calculations. It is a form of deep learning, and convolutional neural networks have characterization learning characteristics, which can classify information into different data. Remote Sensing Data Processing from various sensors has been attracting with more information in Remote Sensing. Remote sensing data is generally adjusted and refined through image processing. Image processing techniques, such as filtering and feature detection, are ideal for dealing with the high-dimensionality of geographically distributed systems. The geological entity is a term in geological work which refers to the product of geological processes that occupy a certain space in the Earth's crust and are different from other materials. They are of different sizes and are divided into different types according to their size. It mainly focuses on improving classification accuracy and accurately describing scattering types. For geological entity recognition, this paper proposed a Deep Convolutional Neural Network Polarized Synthetic Aperture Radar (DCNN-PSAR). It is expected to use deep convolutional neural network technology and polarized SAR technology to explore new methods of geological entities and improve geological recognition capabilities. With the help of Multimodal Remote Sensing Data Processing, it is now possible to characterize and identify the composition of the Earth's surface from orbital and aerial platforms. This paper proposes a ground object classification algorithm for polarized SAR images based on a fully convolutional network, which realizes the geological classification function and overcomes the shortcomings of too long. The evaluation of DCNN-PSAR shows that the accuracy of the water area is showing a rising trend, and the growth rate is relatively fast in the early stage, which directly changes from 0.14 to 0.6. Still, the increase is slower in the later stage. DCNN-PSAR achieves the highest quality of remote sensing data extraction. [ABSTRACT FROM AUTHOR]

Breast cancer is a deadly disease which is most commonly diagnosed in women. It spreads worldwide as the number of cases is increasing each day. The significant increase in cancer cases leads researchers to develop imaging tools for its detection. However, false-positive rates in manual detections are more, which may be due to human error, time taking process or some other issues. If these cancers are not detected at an early stage, they can cause the death of the patient. Therefore, several machine learning and deep learning-based methodologies have come into existence, which can be used to detect breast cancer in early stages. Nowadays, deep learning-based methods are trending, therefore, this paper discusses some important work done by researchers using deep learning. This survey provides an in-depth study of various deep learning models used for breast cancer detection. The details about mammography, histopathology and ultrasound datasets are also given which are being used for comparative analysis of the developed methods. This survey is not only on one imaging modality instead covering three modalities and thus giving a wider study of various methods. Important works from various years have been discussed along with recent works based on vision transformer. Therefore, this paper presents all the necessary information to researchers working in this area and they can think of new ideas based on it to carry out further work. [ABSTRACT FROM AUTHOR]

The use of deep learning (DL) techniques for side-channel analysis (SCA) has become increasingly popular recently. This paper assesses the application of DL to non-profiled SCA attacks on AES-128 encryption, taking into consideration various challenges, including high-dimensional data, imbalanced classes, and countermeasures. The paper proposes using a multi-layer perceptron (MLP) and a convolutional neural network (CNN) to tackle hiding protection methods, such as noise generation and de-synchronization. The paper also introduces a technique called significant Hamming weight (SHW) labeling and a dataset reconstruction approach to handle imbalanced datasets, resulting in a reduction of 30% in the number of measurements required for training. The experimental results on reconstructed dataset demonstrate improved performance in DL-based SCA compared to binary labeling techniques, especially in the face of hiding countermeasures. This leads to better results for non-profiled attacks on different targets, such as ASCAD and RISC-V microcontrollers. [ABSTRACT FROM AUTHOR]

At present, wearable devices have some problems, such as poor adaptability to human motion behavior, and the recognition accuracy required for different wearers cannot be achieved. Based on the principle of deep learning algorithm, this paper realizes the development of intelligent rope skipping movement data monitoring system. Through the universal human body analysis model, the attention mechanism is introduced and embedded into the decoding network. The data set of rope skipping is classified by multiple labels, and the convolution of spatial graph is constructed, which is extended to the time series dynamics of moving human skeleton data. Aiming at the problem of complex information data in the process of moving human body recognition, we use pose estimation to calculate the key points of moving human body, extract the dynamic structure information of human skeleton sequence. Due to the problems of line of sight occlusion in the process of moving human target tracking, a target tracking algorithm based on multi domain convolution neural network is adopted to improve the feature extraction ability of the algorithm by segmenting the target to be tracked and identifying the area around the target. The data set of rope skipping is collected by wearable sensors, and the difference in the numerical range may be large, so the data is normalized. Finally, through the loss function, the fitting effect of neural network can be evaluated, and the gradient optimization model parameters can be calculated, and coping with different data changes. Through the final system performance test, it is verified that the accuracy rate of the system designed in this paper is above 90%, which can effectively monitor the data of skipping rope and be used in the actual operation of skipping rope. [ABSTRACT FROM AUTHOR]

Modern web application development involves handling enormous amounts of sensitive and consequential data. Security is, therefore, a crucial component of developing web applications. A web application's security is concerned with safeguarding the data it processes. The web application framework must have safeguards to stop and find application vulnerabilities. Among all web application attacks, SQL injection and XSS attacks are common, which may lead to severe damage to Web application data or web functionalities. Currently, there are many solutions provided by various study for SQLi and XSS attack detection, but most of the work shown have used either SQL/XSS payload-based detection or HTTP request-based detection. Few solutions available can detect SQLi and XSS attacks, but these methods provide very high false positive rates, and the accuracy of these models can further be improved. We proposed a novel approach for securing web applications from both cross-site scripting attacks and SQL injection attacks using decoding and standardization of SQL and XSS payloads and HTTP requests and trained our model using hybrid deep learning networks in this paper. The proposed hybrid DL model combines the strengths of CNNs in extracting features from input data and LSTMs in capturing temporal dependencies in sequential data. The soundness of our approach lies in the use of deep learning techniques that can identify subtle patterns in the data that traditional machine learning-based methods might miss. We have created a testbed dataset of Normal and SQLi/XSS HTTP requests and evaluated the performance of our model on this dataset. We have also trained and evaluated the proposed model on the Benchmark dataset HTTP CSIC 2010 and another SQL/XSS payload dataset. The experimental findings show that our proposed approach effectively identifies these attacks with high accuracy and a low percentage of false positives. Additionally, our model performed better than traditional machine learning-based methods. This soundness approach can be applied to various network security applications such as intrusion detection systems and web application firewalls. Using our model, we achieved an accuracy of 99.84%, 99.23% and 99.77% on the SQL-XSS Payload dataset, Testbed dataset and HTTP CSIC 2010 dataset, respectively. [ABSTRACT FROM AUTHOR]

In this paper, we introduce misclassification information for the improved training of convolutional neural network classifiers (CNNCs) for image recognition. We construct an additional autoencoder neural network, called tutor, that forces the CNNCs to learn the difference between the misclassified picture and the picture corresponding to the misclassified category. Making full use of the classification results to guide the CNNCs for purposeful learning is expected to improve learning efficiency and classification performance. We integrate the proposed tutor into several state-of-the-art CNNCs architectures and demonstrate improvement in their recognition performance on CIFAR-10/100 and MNIST datasets. Our results suggest that making the most of misclassification information to guide the training of the model can lead to significant performance improvement. [ABSTRACT FROM AUTHOR]

Deep learning techniques have gained immense popularity recently because of their remarkable capacity to learn complex patterns and features from large datasets. These techniques have revolutionized many fields by achieving advanced performance in various tasks. The availability of large datasets and the advancement of computing resources have enabled deep learning models to perform well in solving challenging problems. As a result, they have become an essential tool in many industries, including agriculture. The application of deep learning in agriculture has great potential for increasing productivity, reducing costs, and improving sustainability by aiding in the early identification and prevention of plant leaf diseases, optimizing crop yields, and facilitating precision agriculture. This paper suggests using a novel approach to automatically classify multi-class leaf diseases in tomatoes using a deep multi-scale convolutional neural network (DMCNN). The proposed DMCNN architecture consists of parallel streams of convolutional neural networks at different scales, which get merged at the end to form a single output. The images of tomato leaves are preprocessed using data augmentation techniques and fed into the DMCNN model to classify disease. The proposed approach is evaluated on a dataset of tomato plant images containing 10 distinct classes of diseases and compared with different existing models. The research results reveal that the suggested DMCNN model performs better than other models in terms of accuracy, precision, recall, and F1 score. Furthermore, the proposed model reported an overall accuracy of 99.1%, which is higher than the accuracy of existing models tested on the same dataset. The study demonstrates the potential of deep learning techniques for automated disease classification in agriculture, which can aid in early disease detection and prevent crop loss. [ABSTRACT FROM AUTHOR]

Conventionally, fine-tuning or transfer learning using a pre-trained convolutional network is adopted to design a classifier. However, when the dataset is small this can deteriorate the classifier generalization performance due to negative transfer or overfitting issues. In this paper, we suggest a flexible and high-capacity multiple kernel-based convolutional neural network (MK-CNN) to automate the pathological brain classification task. The proposed network employed different stacks of convolution with various kernels to obtain multi-scale features from the input image. The smaller kernel size provides specific information about the local features whereas the larger kernel size provides the global spatial information. Hence, the network takes into account both regional specifics and global spatial consistency thanks to this multi-scale methodology. Only the output layer is shared between each network stack. This makes it possible to specifically tweak the CNN's weights and biases for each convolution stack and associated kernel size. The results reported on real patient data from the Harvard Whole Brain Atlas reveal that our method outperforms state-of-the-art techniques. The suggested approach may be used to help experts carry out the clinical follow-up study. [ABSTRACT FROM AUTHOR]

Traffic time series anomaly detection has been intensively studied for years because of its potential applications in intelligent transportation. However, classical traffic anomaly detection methods often overlook the evolving dynamic associations between road network nodes, which leads to challenges in capturing the long-term temporal correlations, spatial characteristics, and abnormal node behaviors in datasets with high periodicity and trends, such as morning peak travel periods. In this paper, we propose a mirror temporal graph autoencoder (MTGAE) framework to explore anomalies and capture unseen nodes and the spatiotemporal correlation between nodes in the traffic network. Specifically, we propose the mirror temporal convolutional module to enhance feature extraction capabilities and capture hidden node-to-node features in the traffic network. Morever, we propose the graph convolutional gate recurrent unit cell (GCGRU CELL) module. This module uses Gaussian kernel functions to map data into a high-dimensional space, and enables the identification of anomalous information and potential anomalies within the complex interdependencies of the traffic network, based on prior knowledge and input data. We compared our work with several other advanced deep-learning anomaly detection models. Experimental results on the NYC dataset illustrate that our model works best compared to other models for traffic anomaly detection. [ABSTRACT FROM AUTHOR]

At present times, indoor surveillance becomes a hot research topic among researchers and business sectors. Human detection is one of the vital areas of focus in the surveillance system owing to its significance in proper person detection, human activity identification, and scene classification. Since the indoor spaces comprise poor lighting, variable illuminations, shadowing, and complex background, the human detection process becomes a tedious task. The advent of computer vision and deep learning (DL) models is commonly employed for human detection. This article presents a new intelligent deep learning model for human detection in indoor surveillance videos (IDL-HDIS). As data augmentation process is one of the most renowned ways to increase the size of the dataset which is highly essential for enhancing the prediction accuracy of the model, the same is carried out as a part of even this research work which includes performing rotation, translation and flipping. The IDL-GDIS model uses Faster Region Convolutional Neural Network (Faster R-CNN) model for human detection. The Faster R-CNN comprises of Fast R-CNN and Region Proposal Network (RPN). The RPN uses Capsule Networks (CapsNet) model as a shared convolution neural network (CNN), which acts as a feature extractor and generates the feature map. Besides, dropout is employed to avoid overfitting problem in the CapsNet architecture, the validation of IDL-HDIS model is done by a comprehensive simulation analysis under different aspects. The validation is supported by the evident results of the IDL-HDIS model which is given in the paper. [ABSTRACT FROM AUTHOR]

Adolescent mental health issues like depression and anxiety are rising among adolescents worldwide, leading to serious long-term consequences if left untreated. However, stigma and lack of resources delay access to timely screening and care. Traditional subjective methods fail to detect early warning signs and risk factors. This paper develops a novel hybrid convolutional neural network (CNN) and long short-term memory (LSTM) model to enable early warning of common mental health risks like depression in teenagers. The model is trained on a large clinical dataset with over 50,000 adolescents encompassing electronic health records and neuroimaging data. This allows discovering subtle predictive patterns within clinical encounters and brain structure or function at the population level. The methodology centers on a tailored CNN and LSTM architecture to comprehensively model the multivariate clinical time series data, combining CNN spatial feature learning with LSTM temporal sequence modeling. The hybrid approach leverages complementary techniques to analyze the neuroimaging and electronic health record data. The model achieves an accuracy of 95%, AUC of 97%, precision of 94%, recall of 91%, and F1 score of 92% on held-out test data, significantly outperforming prior state-of-the-art models which had lower accuracy. Detailed predictive feature analysis and clinical validation confirm the model's utility for early mental health risk screening and targeted intervention in teenagers. Overall, the study demonstrates an effective approach to integrating big data, deep learning and rigorous evaluation for developing more accurate, explainable and useful models to support adolescent mental healthcare through data-driven early warning systems. [ABSTRACT FROM AUTHOR]

The widespread availability of digitized fine art collections in museums and galleries has generated a demand for efficient software tools. These tools enable rapid retrieval and semantic classification of art images. Traditional image classification methods often rooted in shallow structure learning algorithms offer the capacity to extract various image features. However, there is need for deep understanding of fundamental painting knowledge during this process due to the reason that certain features may be lost. In this paper, we present an innovating approach for classifying digital paintings based on artist attribution. Our approach centers on the creations of multi-scale pyramid representation derived from a given painting image which facilitate the incorporation of both global and local information within a single image. The training of Convolutional Neural Network (CNN) algorithm is performed for the assignment of class labels on each level of the pyramid. To establish connections among localized image patches, we use Markov Random Fields through the optimization of the Gibbs energy function. We validate the proposed approach which referred to as the multi-scale CNN framework using fusion-based Markov random field using two challenging painting image datasets: WikiArt and Web Gallery of Art (WGA). The proposed model archives better results using WikiArt dataset for painting image classification, i.e., precision (74.12%), recall (70.23%), F1-score (72.65%), and accuracy (80.00%). Using Web Gallery of Art (WGA) dataset for painting image classification, the model achieves the following results: precision (75.88%), recall (77.25%), F1-score (76.44%), and accuracy (78.00%). Our proposed method outperforms state-of-the-art methods excelling not only in image classification but also in terms of computational efficiency. [ABSTRACT FROM AUTHOR]

Identifying a scene based on the environment in which the related audio is recorded is known as acoustic scene classification (ASC). In this paper, a bi-level light-weight Convolutional Neural Network (CNN)-based model is presented to perform ASC. The proposed approach performs classification in two levels. The scenes are classified into three broad categories in the first level as indoor, outdoor, and transportation scenes. The three classes are further categorized into individual scenes in the second level. The proposed approach is implemented using three features: log Mel band energies, harmonic spectrograms and percussive spectrograms. To perform the classification, three CNN classifiers, namely, MobileNetV2, Squeeze-and-Excitation Net (SENet), and a combination of these two architectures, known as SE-MobileNet are used. The proposed combined model encashes the advantages of both MobileNetV2 and SENet architectures. Extensive experiments are conducted on DCASE 2020 (IEEE AASP Challenge on Detection and Classification of Acoustic Scenes and Events) Task 1B development and DCASE 2016 ASC datasets. The proposed SE-MobileNet model resulted in a classification accuracy of 96.9% and 86.6% for the first and second levels, respectively, on DCASE 2020 dataset, and 97.6% and 88.4%, respectively, on DCASE 2016 dataset. The proposed model is reported to be better in terms of both complexity and accuracy as compared to the state-of-the-art low-complexity ASC systems. [ABSTRACT FROM AUTHOR]

Dry eye disease (DED) is a chronic eye disease and a common complication among the world's population. Evaporation of moisture from tear film or a decrease in tear production leads to an unstable tear film which causes DED. The tear film breakup time (TBUT) test is a common clinical test used to diagnose DED. In this test, DED is diagnosed by measuring the time at which the first breakup pattern appears on the tear film. TBUT test is subjective, labour-intensive and time-consuming. These weaknesses make a computer-aided diagnosis of DED highly desirable. The existing computer-aided DED detection techniques use expensive instruments for image acquisition which may not be available in all eye clinics. Moreover, among these techniques, TBUT-based DED detection techniques are limited to finding only tear film breakup area/time and do not identify the severity of DED, which can essentially be helpful to ophthalmologists in prescribing the right treatment. Additionally, a few challenges in developing a DED detection approach are less illuminated video, constant blinking of eyes in the videos, blurred video, and lack of public datasets. This paper presents a novel TBUT-based DED detection approach that detects the presence/absence of DED from TBUT video. In addition, the proposed approach accurately identifies the severity level of DED and further categorizes it as normal, moderate or severe based on the TBUT. The proposed approach exhibits high performance in classifying TBUT frames, detecting DED, and severity grading of TBUT video with an accuracy of 83%. Also, the correlation computed between the proposed approach and the Ophthalmologist's opinion is 90%, which reflects the noteworthy contribution of our proposed approach. [ABSTRACT FROM AUTHOR]

Due to the periodic changes of climate, cracks are widely developed in expansive soils, leading to the destruction of soil integrity, the deterioration of physical strength, and eventually the instability of the expansive soil slope and other disasters. In this paper, a deep learning-based modeling method was proposed for soil crack networks characterization through steps of segmentation, quantification and simulation. Inspired by the U-Net convolutional neural network, a dilated convolution module was added to the backbone to enhance the crack segmentation capability and a subpixel edge detection algorithm was followed for accurate crack edge detection. Then, a deterministic or stochastic method was designed for crack network simulation. A case study of an expansive soil slope on the bank of Wadong Main Canal in the Pi-Shi-Hang Irrigation District, China, was conducted. Results show that the dilated U-Net model gained 0.24 and 0.25 improvement in F1-score and IoU (Intersection over Union) comparing to the conventional segmentation method (Otsu) and crack edge precision was further improved by 5.38%. The performances of proposed method including the image labeling, effect of crack thickness and environmental conditions, etc., are also explored. To validate the simulated crack network, crack areas at different depths were measured through X-ray images. Comparing with the simulation result, mean error rate of 1.05% was achieved. Thus, the proposed method can efficiently characterize soil crack networks using the field acquired crack images. [ABSTRACT FROM AUTHOR]

Handwritten character recognition (HCR) is a challenging task because of high intra-class dissimilarity and high inter-class similarity among the character images. Therefore, the features that are used to represent the character images should be such that they maximize the inter-class variations and minimize intra-class variations. In this paper, a multilevel approach for recognition of Meitei Mayek handwritten characters using fusion of features strategy is presented. The proposed methodology has two levels of recognition. The first level employs a simple CNN. A method to identify the structurally similar character classes based on the probability values of the softmax function is also proposed. The fused feature set is employed in the second-level recognition to distinguish between these character classes, thereby enhancing the discriminative power of the fused feature set by reducing the search space in the second level. The fused feature set is a combination of traditional handcrafted feature descriptors and deep features learned by CNN. The fused feature set is fed to an SVM for second-level recognition. The experimental results show a superior performance of the proposed methodology in identifying the structurally similar character classes. This achieves an overall recognition accuracy of 97.08%, which sets the benchmark on the concerned Meitei Mayek handwritten character dataset. The methodology also shows significant results on MNIST, DIDA and CArDIS datasets. [ABSTRACT FROM AUTHOR]

Diverse loads frequently damage concrete when it is in use. However, it might be challenging to immediately identify the stress and damage of concrete in actual engineering situations. In order to predict the stress and damage of concrete, a deep learning (DL) model based on the convolutional neural network (CNN) is proposed in this paper. To provide the training and validation data, a finite element (FE) model of uniaxial compression of concrete specimens based on the concrete damage-plasticity (CDP) model is constructed. The DL model is trained with the strain contours with a specified range provided by the FE model as the inputs and the stress and damage assessment of concrete as the outputs. The prediction of the stress and damage of concrete materials was effectively realized by the trained DL model, and it was verified in a larger range of working conditions distinct from the training and verification sets. The results show that the DL algorithm has good accuracy and reliability. By efficiently and correctly recreating the FE prediction results, the DL model offers a method for promptly evaluating the stress and damage of concrete structures under complicated stress circumstances in actual engineering. [ABSTRACT FROM AUTHOR]

In big data era, multi-source heterogeneous data become the biggest obstacle to data sharing due to its high dimension and inconsistent structure. Using text classification to solve the ontology construction and mapping problem of multi-source heterogeneous data can not only reduce manual operation, but also improve the accuracy and efficiency. This paper proposes an ontology construction and mapping scheme based on hybrid neural network and autoencoder. Firstly, the proposed text classification method uses the multi-core convolutional neural network to capture local features and uses the improved Bidirectional Long Short-Term Memory network to compensate for the shortcomings of the convolutional neural network that cannot obtain context-related information. Secondly, a similarity matching method is used for ontology mapping, which integrate autoencoder to improve anti-interference ability. We have carried out several sets of experiments to test the validity of the proposed ontology construction and mapping scheme. [ABSTRACT FROM AUTHOR]

Ovarian cancer is one of the most common types of cancer in women. Correct differentiation between benign and malignant ovarian tumors is of immense importance in medical fields. In this paper, we introduce group penalized logistic regressions to enhance diagnosis accuracy. Firstly, we divide 349 ovarian cancer patients into two sets: one for learning model parameters, and the other for assessing prediction performance, and select 46 variables from 49 traits as the predictor vector to construct GLASSO/GSCAD/GMCP penalized logistic regressions with 11 groups. Secondly, we develop group coordinate descent (GCD) algorithm and its specific pseudo code to simultaneously complete group selection and group estimation, introduce the tenfold cross validation (CV) procedure to select the relatively optimal tuning parameter, and apply the testing set and Youden index to obtain class probability estimator and class index information. Finally, we compute the accuracy, precision, specificity, sensitivity, F1-score and the area under ROC curve (AUC) to assess the prediction performance to the proposed group penalized methods, and found that GLASSO/GSCAD/GMCP penalized logistic regressions outperform three machine learning methods (ANN, artificial neural network; SVM, support vector machine; XGBoost, eXtreme gradient boosting) and three deep learning methods (CNN, convolutional neural network; DNN, deep neural network; RNN, recurrent neural network) in terms of accuracy, F1-score and AUC. [ABSTRACT FROM AUTHOR]

Nowadays, the issue of fluctuations in the price of digital Bitcoin currency has a striking impact on the profit or loss of people, international relations, and trade. Accordingly, designing a model that can take into account the various significant factors for predicting the Bitcoin price with the highest accuracy is essential. Hence, the current paper uses market sentiment and multiple feature extraction to present several Bitcoin price prediction models based on convolutional neural network and long short-term memory. In the proposed models, several parameters, including Twitter data, news headlines, news content, Google Trends, Bitcoin-based stock, and finance, are employed based on deep learning to make a more accurate prediction. Such parameters are the input data used to predict the Bitcoin price. Besides, the proposed model analyzes the Valence Aware Dictionary and Sentiment Reasoner sentiments to examine the market's latest news and cryptocurrencies. According to this study's various inputs and analyses, several effective feature selection methods, including mutual information regression, linear regression, correlation-based, and a combination of the feature selection models, are exploited to predict the price of Bitcoin. Finally, a careful comparison is made between the proposed models in terms of some performance criteria like mean square error (MSE), root-mean-square error, mean absolute error, median absolute error, and coefficient of determination (R2). The obtained results indicate that the proposed hybrid model based on sentiments analysis and combined feature selection with MSE value of 0.001 and R2 value of 0.98 provides better estimations with more minor errors regarding Bitcoin price. This proposed model can also be employed as an individual assistant for more informed trading decisions associated with Bitcoin. [ABSTRACT FROM AUTHOR]

Combining convolutional neural networks (CNNs) and recurrent neural networks (RNNs) produces a powerful architecture for video classification problems as spatial–temporal information can be processed simultaneously and effectively. Using transfer learning, this paper presents a comparative study to investigate how temporal information can be utilized to improve the performance of video classification when CNNs and RNNs are combined in various architectures. To enhance the performance of the identified architecture for effective combination of CNN and RNN, a novel action template-based keyframe extraction method is proposed by identifying the informative region of each frame and selecting keyframes based on the similarity between those regions. Extensive experiments on KTH and UCF-101 datasets with ConvLSTM-based video classifiers have been conducted. Experimental results are evaluated using one-way analysis of variance, which reveals the effectiveness of the proposed keyframe extraction method in the sense that it can significantly improve video classification accuracy. [ABSTRACT FROM AUTHOR]

Deep learning methods have relevant applications in crack detection in buildings. However, one of the challenges in this field is the hyperparameter tuning process for convolutional neural networks (CNN). Thus, the objective of this paper is to propose a automated hyperparameter tuning approach for crack image classification. For this, a public dataset with 40,000 images of walls and floors of several buildings was used. The images are divided into two classes: negative (non-crack) and positive (crack). In this aspect, statistical methods are used for hyperparameter tuning, such as analysis of variance, Scott–Knott method and HyperTuningSK algorithm. Moreover, three new automated machine learning algorithms are proposed: AutoHyperTuningSK, AutoHyperTuningSK-test and AutoHyperTu-ningSK-DA. CNN architecture from the literature (MobileNet) and three types of hyperparameters (learning rate, optimizer and data augmentation) are analyzed. In general, the recommended configurations reached the best results in relation to unselected hyperparameters. In this regard, a selected combinations achieved a mean accuracy of around 99 % (test experiments) in binary crack classification. [ABSTRACT FROM AUTHOR]

Speech enhancement is an important method for improving speech quality and intelligibility in noisy environments. An effective speech enhancement model depends on precise modelling of the long-range dependencies of noisy speech. Several recent studies have examined ways to enhance speech by capturing the long-term contextual information. For speech enhancement, the time–frequency (T–F) distribution of speech spectral components is also important, but is usually ignored in these studies. The multi-stage learning method is an effective way to integrate various deep learning modules at the same time. The benefit of multi-stage training is that the optimization target can be iteratively updated stage by stage. In this paper, speech enhancement is investigated by multi-stage learning using a multi-stage structure in which time–frequency attention (TFA) blocks are followed by stacks of squeezed temporal convolutional networks (S-TCN) with exponentially increasing dilation rates. To reinject original information into later stages, a feature fusion block (FB) is inserted at the input of later stages to reduce the possibility of speech information being lost in the early stages. The S-TCN blocks are responsible for temporal sequence modelling tasks. The time–frequency attention (TFA) is a simple but effective network module that explicitly exploits position information to generate a 2D attention map to characterize the salient T–F distribution of speech by using two branches, time-frame attention and frequency attention in parallel. Extensive experiments have demonstrated that the proposed model consistently improves the performance over existing baselines across two widely used objective metrics such as PESQ and STOI. A significant improvement in system robustness to noise is also shown by our evaluation results using the TFA module. [ABSTRACT FROM AUTHOR]

The K-complex is one of the most important and noticeable features in the electroencephalography (EEG) signal, therefore its detection is critical for EEG signal analysis. It is used to diagnose neurophysiologic and cognitive disorders as well as sleep studies. Existing detection methods largely depend on tedious, time-consuming, and error-prone manual inspection of the EEG waveform. In this paper, a highly accurate K-complex detection system is developed. Based on multiple convolutional neural network (CNN) feature extraction backbones and EEG waveform images, you only look once v3 (YOLOv3) detector was designed, trained, and tested. Extensive performance evaluation was performed using five deep transfer learning feature extraction models; Darknet-53, MobileNets, ResNet-18, SqueezeNet, and Darknet-53-coco. The dataset was comprised of 10948 images of EEG waveforms, with the K-complex location automatically annotated with bounding boxes. The Darknet-53 model performed consistently high (i.e., 89.84–99.44% precision and 10.41–0.55% miss rate). Thus, it is possible to perform automatic K-complex detection in real-time with high accuracy that aid practitioners in speedy EEG inspection. [ABSTRACT FROM AUTHOR]

Face recognition in visual surveillance systems is important for various applications to identify individuals who are behaving defiantly at the time of an event or for investigation purposes. Despite the dramatic improvements in facial recognition technology in recent years, it is difficult to recognize faces from surveillance feeds due to the presence of multiple people of different scales and orientations. This paper solves the task of low-resolution face recognition by combining exemplary techniques for extracting distinct features. This research utilizes the attributes learned by customized and learnable filters and injected in the training process to better match them with human brain functionality. The Gabor transform aims to convolve a facial image using a range of Gabor filter coefficients at various scales and orientations, resulting in scale and rotation invariant features. The tailored architecture with residual stream aims to enhance functional representation and prevent the gradient of the prediction engine from affecting the backbone network functional map. Experimental analysis is performed on the SCface and TinyFace databases and is reported with an accuracy of 89.21% on the SCface database and 56.68% on the TinyFace database. [ABSTRACT FROM AUTHOR]

The contamination of insulators increases their surface conductivity, resulting in a higher chance of shutdowns occurring. To measure contamination, equivalent salt deposit density (ESDD) and non-soluble deposit density (NSDD) are used. In this paper, the VGG-11, VGG-13, VGG-16, VGG-19, ResNet-18, ResNet-34, ResNet-50, ResNet-152, DenseNet-121, DenseNet-161, DenseNet-169, and DenseNet-201 convolutional neural networks (CNNs) were considered to classify the visible contamination of pin-type distribution power grid insulators. The NSDD presents more visual variation than ESDD when artificial contamination is evaluated. Comparing the CNNs, the ResNet-50 had the best performance for classifying visible contamination using unbalanced data with an accuracy of 99.242% and an F1-score of 0.97436, respectively. In benchmarking, the ResNet-50 outperformed well-established classifiers such as the multilayer perceptron, support vector machine, k-nearest neighbors, decision tree, ensemble bagged trees, and quadratic discriminant. [ABSTRACT FROM AUTHOR]

Chest X-ray (CXR) is a prominent and cost-effective medical imaging tool used in healthcare sectors. Coronavirus (COVID-19) has recently proliferated around the world, and different waves are returning, posing a serious threat to the global economy and health. Deep learning networks are being used to detect infected patients from CXR images, which is a huge step in making a prompt COVID-19 pre-diagnosis and reducing the workload of medical staff. However, in CXR images, there are complexity in structure and patterns of gray-scale distribution, as well as complex hemidiaphragm borders and details that might interfere with the machine 's and doctor's diagnosis. In particular, this paper developed a new multi-scale parallel CNN (called MSP-CNN) for denoising CXR images and a distinctive COVID-19 application that can improve image quality. The foundation of MSP-CNN is comprised of several fundamental parts are: (a) Extracting spatial information with a multi-scale feature extractor (b) Efficient Spatial Channel Attention(ESCA) without raising the network parameters, more attention to be paid to complex structures and hemidiaphragm edges in CXR images. Extensive testing has shown that our MSP-CNN is better at preserving complex structures of features in CXR images. Rigorous experiments also show that MSP-CNN has a positive influence on CXR image classification and COVID-19 case detection from denoised CXR images. It is envisioned that with reliable accuracy, this method can be introduced for clinical practices in the future. [ABSTRACT FROM AUTHOR]

Objective: The aim of this work was to assemble a large annotated dataset of bitewing radiographs and to use convolutional neural networks to automate the detection of dental caries in bitewing radiographs with human-level performance. Materials and methods: A dataset of 3989 bitewing radiographs was created, and 7257 carious lesions were annotated using minimal bounding boxes. The dataset was then divided into 3 parts for the training (70%), validation (15%), and testing (15%) of multiple object detection convolutional neural networks (CNN). The tested CNN architectures included YOLOv5, Faster R-CNN, RetinaNet, and EfficientDet. To further improve the detection performance, model ensembling was used, and nested predictions were removed during post-processing. The models were compared in terms of the F 1 score and average precision (AP) with various thresholds of the intersection over union (IoU). Results: The twelve tested architectures had F 1 scores of 0.72–0.76. Their performance was improved by ensembling which increased the F 1 score to 0.79–0.80. The best-performing ensemble detected caries with the precision of 0.83, recall of 0.77, F 1 = 0.80 , and AP of 0.86 at IoU=0.5. Small carious lesions were predicted with slightly lower accuracy (AP 0.82) than medium or large lesions (AP 0.88). Conclusions: The trained ensemble of object detection CNNs detected caries with satisfactory accuracy and performed at least as well as experienced dentists (see companion paper, Part II). The performance on small lesions was likely limited by inconsistencies in the training dataset. Clinical significance: Caries can be automatically detected using convolutional neural networks. However, detecting incipient carious lesions remains challenging. [ABSTRACT FROM AUTHOR]

Deep learning has been acknowledged as an increasingly important technology for ENSO forecasts. The most cutting-edge deep learning algorithm is developed based on Convolutional Neural Network (CNN), which can achieve a multi-year (about 17-month-lead) forecast and has conquered the 'spring forecast barrier' problem. However, this group of methods are still challenged by several critical issues. First, they usually utilize the global sea surface temperature (SST) fields as inputs without considering the specific contributions of variant oceanic regions in ENSO forecasts. Consequently, they cannot effectively investigate the role of the 'teleconnection' mechanism among different oceans (Indian, Pacific, and Atlantic Oceans) and different ocean parts (the tropic and non-tropic regions) especially in the forecast of extreme ENSO events. Second, existing methods mainly utilize the discrete monthly SST fields for ENSO forecasts without investigating the rate-of-changes between adjacent months, which also provides important information to the prediction of variation tendency. To solve these problems, this paper develops a Tendency-and-Attention-Informed Deep Residual Network (TA-DRN) for multi-year ENSO forecasts. The contributions of different oceanic regions can be learned by a spatial attention module while the variation tendency of adjacent previous and current months can be interpreted by the first-and-second order of differences of SST fields. Through informed by these two modules, the performance of TA-DRN can be improved significantly, especially in predicting extreme El Niño and La Niña events. [ABSTRACT FROM AUTHOR]

Soybean can be easily contaminated by Aspergillus flavus which can generate toxigenic and endanger human life and health. Due to the difficulty in detecting moldy phenomena at early stage by the naked eye and traditional machine vision technique, this paper proposes a classification method based on deep learning and optical coherence (OCT) techniques to detect moldy phenomenon of soybeans at early stage. The proposed method mainly includes three stages: the first stage is mildew information extraction, we use convolutional neural network (CNN) to extract image features. The input of traditional CNN is usually the whole image, and the output can not to reflect the fine-grained information. On this basis, we use the features extracted from the patch for the perception of fine-grained information (such as tiny mildew pixels). In the second stage, the features of the two channels are fused using the self-attention mechanism. In the third stage, the fused feature vectors containing the region information of moldy spots are used for classification. The results show that the proposed method is superior to the traditional CNN model in early mildew identification, with an average accuracy of 99.5% and have 15 points increasing to traditional CNN model, which proves the effectiveness of the method. [ABSTRACT FROM AUTHOR]

Non-Small cell lung cancer (NSCLC) is one of the most dangerous cancers, with 85% of all new lung cancer diagnoses and a 30–55% of recurrence rate after surgery. Thus, an accurate prediction of recurrence risk in NSCLC patients during diagnosis could be essential to drive targeted therapies preventing either overtreatment or undertreatment of cancer patients. The radiomic analysis of CT images has already shown great potential in solving this task; specifically, Convolutional Neural Networks (CNNs) have already been proposed providing good performances. Recently, Vision Transformers (ViTs) have been introduced, reaching comparable and even better performances than traditional CNNs in image classification. The aim of the proposed paper was to compare the performances of different state-of-the-art deep learning algorithms to predict cancer recurrence in NSCLC patients. In this work, using a public database of 144 patients, we implemented a transfer learning approach, involving different Transformers architectures like pre-trained ViTs, pre-trained Pyramid Vision Transformers, and pre-trained Swin Transformers to predict the recurrence of NSCLC patients from CT images, comparing their performances with state-of-the-art CNNs. Although, the best performances in this study are reached via CNNs with AUC, Accuracy, Sensitivity, Specificity, and Precision equal to 0.91, 0.89, 0.85, 0.90, and 0.78, respectively, Transformer architectures reach comparable ones with AUC, Accuracy, Sensitivity, Specificity, and Precision equal to 0.90, 0.86, 0.81, 0.89, and 0.75, respectively. Based on our preliminary experimental results, it appears that Transformers architectures do not add improvements in terms of predictive performance to the addressed problem. [ABSTRACT FROM AUTHOR]

Automated multi-organ segmentation plays an essential part in the computer-aided diagnostic (CAD) of chest X-ray fluoroscopy. However, developing a CAD system for the anatomical structure segmentation remains challenging due to several indistinct structures, variations in the anatomical structure shape among different individuals, the presence of medical tools, such as pacemakers and catheters, and various artifacts in the chest radiographic images. In this paper, we propose a robust deep learning segmentation framework for the anatomical structure in chest radiographs that utilizes a dual encoder–decoder convolutional neural network (CNN). The first network in the dual encoder–decoder structure effectively utilizes a pre-trained VGG19 as an encoder for the segmentation task. The pre-trained encoder output is fed into the squeeze-and-excitation (SE) to boost the network's representation power, which enables it to perform dynamic channel-wise feature calibrations. The calibrated features are efficiently passed into the first decoder to generate the mask. We integrated the generated mask with the input image and passed it through a second encoder–decoder network with the recurrent residual blocks and an attention the gate module to capture the additional contextual features and improve the segmentation of the smaller regions. Three public chest X-ray datasets are used to evaluate the proposed method for multi-organs segmentation, such as the heart, lungs, and clavicles, and single-organ segmentation, which include only lungs. The results from the experiment show that our proposed technique outperformed the existing multi-class and single-class segmentation methods. [ABSTRACT FROM AUTHOR]

Surface-enhanced Raman spectroscopy (SERS), as a rapid, non-invasive and reliable spectroscopic detection technique, has promising applications in disease screening and diagnosis. In this paper, an annealed silver nanoparticles/porous silicon Bragg reflector (AgNPs/PSB) composite SERS substrate with high sensitivity and strong stability was prepared by immersion plating and heat treatment using porous silicon Bragg reflector (PSB) as the substrate. The substrate combines the five deep learning algorithms of the improved AlexNet, ResNet, SqueezeNet, temporal convolutional network (TCN) and multiscale fusion convolutional neural network (MCNN). We constructed rapid screening models for patients with primary Sjögren's syndrome (pSS) and healthy controls (HC), diabetic nephropathy patients (DN) and healthy controls (HC), respectively. The results showed that the annealed AgNPs/PSB composite SERS substrates performed well in diagnosing. Among them, the MCNN model had the best classification effect in the two groups of experiments, with an accuracy rate of 94.7% and 92.0%, respectively. Previous studies have indicated that the AgNPs/PSB composite SERS substrate, combined with machine learning algorithms, has achieved promising classification results in disease diagnosis. This study shows that SERS technology based on annealed AgNPs/PSB composite substrate combined with deep learning algorithm has a greater developmental prospect and research value in the early identification and screening of immune diseases and chronic kidney disease, providing reference ideas for non-invasive and rapid clinical medical diagnosis of patients. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) is one of the main causes of blindness in people around the world. Early diagnosis and treatment of DR can be accomplished by organizing large regular screening programs. Still, it is difficult to spot diabetic retinopathy timely because the situation might not indicate signs in the primary stages of the disease. Due to a drastic increase in diabetic patients, there is an urgent need for efficient diabetic retinopathy detecting systems. Auto-encoders, sparse coding, and limited Boltzmann machines were used as a few past deep learning (DL) techniques and features for the classification of DR. Convolutional Neural Networks (CNN) have been identified as a promising solution for detecting and classifying DR. We employ the deep learning capabilities of efficient net batch normalization (BNs) pre-trained models to automatically acquire discriminative features from fundus images. However, we successfully achieved F1 scores above 80% on all efficient net BNs in the EYE-PACS dataset (calculated F1 score for DeepDRiD another dataset) and the results are better than previous studies. In this paper, we improved the accuracy and F1 score of the efficient net BNs pre-trained models on the EYE-PACS dataset by applying a Gaussian Smooth filter and data augmentation transforms. Using our proposed technique, we have achieved F1 scores of 84% and 87% for EYE-PACS and DeepDRiD. [ABSTRACT FROM AUTHOR]

The powerful parallel computing capability of GPU and the development of matrix processing unit in recent years provide more possibilities to improve the performance of convolutional neural network (CNN) on GPU. For the Winograd convolution algorithm, which is the most widely used in CNN and has the best performance, there are already some tuning results, but they all ignore the utilization of the matrix operation unit and fail to make full use of the computing resources of GPU. This paper introduces a single precision accelerated solution on GPU for CNN. According to the indicators of architecture, the optimal data layout, grid division and block division methods are derived. In order to adapt to a variety of padding in practical application, an efficient dynamic scheme for filling is designed, and by the use of matrix cores, a pipeline algorithm with operator fusion is implemented. The deep learning accelerated library MIOpen in AMD is used as the baseline. Taking several convolutional layers of ResNet50 as the experimental input, the evaluation shows that our approach outperforms MIOpen with the speedup of 1.41x on MI210, and reaches 74% of the peak value of single precision calculations. Applying this method to the training and inference of ResNet50, the speedup of 1.68x is obtained. [ABSTRACT FROM AUTHOR]

Based on the deep learning method, a network model that can quickly and accurately identify the species of Fritillaria cirrhosa species was constructed. The learning method based on deep residual convolutional neural network was used to input the unprocessed original image directly as input, and the features of the image were extracted through convolution and pooling operations. On this basis, the ResNet34 model was improved, and the additional fully connected layer was added in front of the Softmax classifier to improve the learning ability of the network model. Total of 3915 images of three kinds of Fritillaria cirrhosa were used as data sources for the experiments, among which 160 images of each type were randomly selected to form the validation set. The final training set recognition accuracy rate was 95.8%, the validation set accuracy rate reached 92.3%, and the test set accuracy rate was 88.7%. The image recognition method of Fritillaria cirrhosa based on deep learning proposed in this paper is effective and feasible, which can quickly and accurately identify the species of Fritillaria cirrhosa species, and provides a new idea for the intelligent recognition of Chinese medicinal materials. [ABSTRACT FROM AUTHOR]

To assist ophthalmologists in diagnosing retinal abnormalities, Computer Aided Diagnosis has played a significant role. In this paper, a particular Convolutional Neural Network based on Wavelet Scattering Transform (WST) is used to detect one to four retinal abnormalities from Optical Coherence Tomography (OCT) images. Predefined wavelet filters in this network decrease the computation complexity and processing time compared to deep learning methods. We use two layers of the WST network to obtain a direct and efficient model. WST generates a sparse representation of the images which is translation-invariant and stable concerning local deformations. Next, a Principal Component Analysis classifies the extracted features. We evaluate the model using four publicly available datasets to have a comprehensive comparison with the literature. The accuracies of classifying the OCT images of the OCTID dataset into two and five classes were 100 % and 82.5 % , respectively. We achieved an accuracy of 96.6 % in detecting Diabetic Macular Edema from Normal ones using the TOPCON device-based dataset. Heidelberg and Duke datasets contain DME, Age-related Macular Degeneration, and Normal classes, in which we achieved accuracy of 97.1 % and 94.4 % , respectively. A comparison of our results with the state-of-the-art models shows that our model outperforms these models for some assessments or achieves nearly the best results reported so far while having a much smaller computational complexity. [ABSTRACT FROM AUTHOR]

With the popularity of video surveillance network, image-based fire detection method is of great significance to reduce the loss of public life and public property caused by fire disasters. Convolutional neural networks based on deep learning have been used in the field of fire detection. However, these methods directly use the existing object detection network, so there are some problems such as low detection accuracy, slow speed and high calculation cost. In this paper, we propose a cost-efficient neural network based on long-short range attention for real-time fire detection. First, we design a light-weight backbone network to extract multi-scale fire features with lower computational cost. Secondly, we introduce the transformer module into the convolution layer and construct an long-short range attention block, which can extract the global attention independent of distance to assist the network in identifying fire and background. Finally, the feature fusion module is constructed to process the multi-scale features extracted by backbone, and improve the detection effect of different size fires, especially early small-size fires. The experimental results show that our network can accurately detect fire with very fast speed and very low calculation cost, and the false alarm rate is lower. At the same time, it also has significant advantages for the detection performance of early small-size fire. [ABSTRACT FROM AUTHOR]

Rock quality designation (RQD), as a well-accepted and appliable rock quality index, is crucial in geotechnical engineering. Current RQD estimation mainly relies on either manual statistics or the image binarisation method, while the former approach surrenders high labour intensity and low efficiency and the latter one is constrained by image acquisition. Considering the above-mentioned limitations in RQD estimation, this study proposed a novel convolutional neural network (CNN) approach to automatically perform core recognition and RQD cataloguing with significant improvement in accuracy and efficiency. Firstly, the proposed neural network automatically identified the prefabricated round markers to distinct drilling rounds. To maximumly strengthen the engineering capability of CNN without losing generality, we considered image inversion, rotation, noise addition, and RGB conversion of 200 core box samples in total. Secondly, replacing the unstable image binarisation method, the advanced YOLO V2 object detection model, a single-stage real-time object detection model, was adopted in this study. We also proposed the modified four-layer downsampling structure as our CNN, and then developed an automatic recognition approach for both cores and the round markers, resulting in a 93.1% accuracy according to the validation set. Thirdly, this study proposed an auto-ranking algorithm to sequence the core sample according to the confidence of core recognition by the CNN and row-scanning results for subsequent RQD cataloguing. In addition, the optimal scan width was proved to be 1.33 times larger than the average core width. Finally, a quick cataloguing platform for drill cores was developed. Compared with manual measurement and visual statistics, intelligent RQD cataloguing is characterised by its unparalleled accuracy and efficiency, which is merited by the low relative error (1.84%) and fast processing time (around 0.2 s). Moreover, the application presented in this paper is applicable to most geotechnical engineering scenarios. This is attributed to its low requirements in image acquisition, high efficiency, precise recognition, and robustness. [ABSTRACT FROM AUTHOR]

Snapshot compressive imaging (SCI) systems aim to capture high-dimensional ( ≥ 3 D) images in a single shot using 2D detectors. SCI devices consist of two main parts: a hardware encoder and a software decoder. The hardware encoder typically consists of an (optical) imaging system designed to capture compressed measurements. The software decoder, on the other hand, refers to a reconstruction algorithm that retrieves the desired high-dimensional signal from those measurements. In this paper, leveraging the idea of deep unrolling, we propose an SCI recovery algorithm, namely GAP-net, which unfolds the generalized alternating projection (GAP) algorithm. At each stage, GAP-net passes its current estimate of the desired signal through a trained convolutional neural network (CNN). The CNN operates as a denoiser projecting the estimate back to the desired signal space. For the GAP-net that employs trained auto-encoder-based denoisers, we prove a probabilistic global convergence result. Finally, we investigate the performance of GAP-net in solving video SCI and spectral SCI problems. In both cases, GAP-net demonstrates competitive performance on both synthetic and real data. In addition to its high accuracy and speed, we show that GAP-net is flexible with respect to signal modulation implying that a trained GAP-net decoder can be applied in different systems. Our code is available at https://github.com/mengziyi64/GAP-net. [ABSTRACT FROM AUTHOR]

With the worldwide spread of the COVID-19 pandemic, the demand for medical syringes has increased dramatically. Scale defect, one of the most common defects on syringes, has become a major barrier to boosting syringe production. Existing methods for scale defect detection suffer from large volumes of data requirements and the inability to handle diverse and uncertain defects. In this paper, we propose a robust scale defects detection method with only negative samples and favorable detection performance to solve this problem. Different from conventional methods that work in a batch-mode defects detection manner, we propose to locate the defects on syringes with a two-stage framework, which consists of two components, that is, the scale extraction network and the scale defect discriminator. Concretely, the SeNet is first built to utilize the convolutional neural network to extract the main structure of the scale. After that, the scale defect discriminator is designed to detect and label the scale defects. To evaluate the performance of our method, we conduct experiments on one real-world syringe dataset. The competitive results, that is, 99.7% on F1, prove the effectiveness of our method. [ABSTRACT FROM AUTHOR]

Automatic and accurate polyp segmentation is significant for diagnosis and treatment of colorectal cancer. This is a challenging task due to the polyp's shape and size diversity. Recently, various deep convolutional neural networks have been developed for polyp segmentation. However, most state-of-the-art methods have suffered from a poor performance in the segmentation of smaller, flat, or noisy polyp objects. In the paper, we propose a novel Deep Context Attention Network (DCANet) for accurate polyp segmentation based on an encoder–decoder framework. ResNet34 is adopted as the encoder, and five functional modules are introduced to improve the performance. Specifically, the improved local context attention module (LCAM) and global context module (GCM) are exploited to efficiently extract the local multi-scale and global multi-receptive-field context information, respectively. Then, an enhanced feature fusion module (FFM) is devised to effectively select and aggregate context features through spatial-channel attention. Finally, equipped with elaborately designed multi-attention modules (MAM), new decoder and supervision blocks are developed to accurately predict polyp regions via powerful channel-spatial-channel attention. Extensive experiments are conducted on the Kvasir-SEG and EndoScene benchmark datasets. The results demonstrate that the proposed network achieves superior performance compared to other state-of-the-art models. The source code will be available at https://github.com/ZAKAUDD/DCANet. [ABSTRACT FROM AUTHOR]

Automatic facial expression recognition plays a crucial role in computer vision, and pattern recognition. However, most existing deep learning-based facial expression classifiers usually obtain high average accuracy but have poor recognition accuracy for difficult expressions, like fear and disgust. In this paper, we propose a novel end-to-end architecture termed two-stream inter-class variation enhancement network, which learns the high-level semantic features and subtle inter-class variations in a joint fashion. More precisely, the global feature extraction network is used to extract spatial-channel semantic features, and the variations between different expressions are modeled by a distinction-reinforced network. The outputs of these two streams are weighted integrated in the expression classification network. In addition, a class balanced-weighted cross-entropy loss is designed to further improve feature discrimination. Experiment results indicate that the proposed network can significantly improve the recognition of difficult expressions and achieve a satisfactory average recognition accuracy of 73.67% on FER2013, 86.17% on RAFDB, 98.19% on CK+, and 98.85% on Oulu-CASIA, which outperforms the other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The automatic reading of pointer meters is significantly important to data monitoring and efficient measurement in the industrial field. However, the existing automatic reading method can not obtain accurate performance in natural scenarios and present no satisfactory application effects in industrial fields (such as power stations and gas stations). In this paper, a novel automatic reading method for pointer meters based on deep learning is proposed, which contains five stages. Stage-1: the object detection algorithm Yolov4 and the feature optimization module IFF are used to locate the target meter. Stage-2: Semantic segmentation model is applied to extract the pointer area based on Anam-Net. Stage-3: the character detection algorithm CRAFT and the text recognition algorithm E2E-MLT are combined and used to recognize the scale text and unit on the meter. Stage-4: the scale area of the meter is converted to the polar coordinate system, and a lightweight convolutional neural network is designed to locate the main scale line. And finally in Stage-5: the reading data are calculated according to the outputs of the above-mentioned deep learning models. The experiment results show that the reading method proposed in this paper has higher accuracy and robustness than those of the existing approaches and obtains satisfactory application effects in the industrial field. [ABSTRACT FROM AUTHOR]