Search

In the field of abstract text summarization, architectures based on encoder-decoder frameworks are widely applied to sequence-to-sequence generation tasks and can effectively handle sequences of unlimited length. Subsequently, the transformer model use a global attention mechanism, allowing encodings at different distances to mutually interact, greatly enhancing the model’s contextual awareness. However, this context-awareness is global, requiring the model to additionally learn to extract different levels of information to increase understanding. We improve the structure of the model to introduce prior knowledge so that it can learn from the global and local information and enhance the model’s understanding ability. This paper proposes global information-aware encoding and local information-aware encoding, which enhance the understanding of documents from coarse-grained and fine-grained perspectives respectively. Global encoding adds an extra feature to the encoder stage and performs attention with the document, generating a global summary encoding of the entire document to guide the generation of the summary content. Local encoding is to perform local convolution on the features extracted by the encoder, use prior knowledge to extract local features of the document and enable the model to quickly extract local detail information. Experiments show that the improved model proposed in this paper has higher rouge scores than the baseline model on the LCSTS and CSL datasets, and also has advantages over some mainstream models. The generated summaries are more accurate and informative. The code is available on github. url: https://github.com/keptupp/A-text-summarization-approach-to-enhance-global-and-local-information-awareness-of-transformer.

A strong and reliable Traffic Sign Detection and Recognition (TSDR) system is essential for the effective deployment of autonomous driving technology. In this field, numerous scholars have conducted extensive research, but current studies only consider TSDR under ideal conditions, neglecting scenarios such as rain, snow, fog, which can cause image blurring. This paper investigates the challenges of TSDR performance degradation caused by five adverse environmental conditions: rain, snow, fog, lens dirt, and lens blur. To overcome the adverse effects of these conditions on TSDR, this paper proposes a Convolutional Neural Network (CNN)-based TSDR method, consisting of three modules: adverse environment classification module, image enhancement module, and traffic sign detection module. The adverse environment classifier, based on the VGG19 architecture, identifies whether the image includes the aforementioned five adverse weather conditions. The image enhancement module named Enhance-Net enhances each of the five adverse environments separately and specifically enhances the traffic sign regions within the image, rather than the entire image area. To increase the speed of the proposed method, the traffic sign detection module utilizes the YOLOv4 framework. The proposed method’s effectiveness is assessed using the CURE-TSD dataset, which includes traffic videos recorded in various adverse environmental conditions. Experimental results demonstrate that under five different levels of adverse environments, the pro-posed method achieves 95.03% accuracy and runs at a rate of 12.79 fps (frames per second). In contrast to the current benchmark, although there is a 2.81% reduction in accuracy resulting from training the proposed method on a subset of the dataset, the speed has increased by 12.03 fps, demonstrating the efficacy of the proposed approach.

Predicting the price of soybean meal futures has always been a hot topic in the financial market. It is influenced by many factors and has complex nonlinear characteristics, which makes it difficult to predict accurately. This paper proposes a prediction model for soybean meal futures price based on CNN-ILSTM-SA, which consists of CNN (Convolutional Neural Network), ILSTM (Improved Long Short-Term Memory), and SA (Self-Attention mechanism). CNN can well perform feature extraction on the input time series data. The ILSTM model presented in this paper removes the output gate of LSTM (Long Short-Term Memory) and improves the forget gate and input gate, thus reducing ILSTM’s complexity, shortening training time, and improving learning ability. SA can get the correlation between different input terms and predicted values, and assign different weight coefficients according to the correlation level to improve prediction accuracy. To verify the CNN-ILSTM-SA’s effectiveness, we compare it with eighteen baseline models, and the experimental results show that it is optimal. The MAE (Mean Absolute Error) of CNN-ILSTM-SA is 39.0441, the smallest of all models. As the data of the soybean meal futures price fluctuates in the range of 3500-4500 in experiments, the proportion of the MAE value of 39.0441 is 0.87%-1.15%, which is at an extremely low level; $\mathrm {R}^{2}$ (R-square) is 0.97487, which is closest to 1; The training time is 157.64s.

Ultra-wide Band (UWB) technology has emerged as a pivotal tool for human motion detection, finding applications in diverse areas ranging from smart homes to automotive safety. This paper presents a comprehensive survey of methodologies employed in UWB-based motion detection, elucidating their strengths, challenges, and performance metrics. While several methods, including Convolutional Neural Network (CNN) approaches, have been explored, challenges such as motion state overlaps, the necessity for enhanced spatial resolution, and background noise interference persist. Among the various methods analyzed, the SGWO-based RMDL technique emerges as a frontrunner, offering superior accuracy, reduced mean squared error, and impressive true negative and positive rates. Moreover, its computational efficiency sets a precedent in human motion detection. This paper provides insights into the state-of-the-art Through the wall imaging and human vital signs observation for future research and realtime applications.

Radar echo image sequence prediction is a spatial-temporal forecasting problem, which is one of the main challenges in precipitation nowcasting. In response to issues such as poor extraction of spatiotemporal features and blurred prediction of image sequence in previous models, this paper proposes a spatial-temporal dual discriminator precipitation nowcasting method using SNGAN with InceptionGRU (STD-SNGAN-InceptionGRU). The design of temporal and spatial dual discriminators constrains the generator’s prediction samples, effectively enhancing the spatiotemporal prediction capability of the generator through adversarial training. The generator utilizes multiscale convolution (Inception) and multiscale gate recurrent unit (InceptionGRU) to enhance its ability to extract spatiotemporal features, thereby improving its learning capability for the evolution of radar echo sequences. The comparative experimental results demonstrate that the proposed STD-SNGAN-InceptionGRU model outperforms other algorithms in terms of critical success index (CSI) and probability of detection (POD) at different echo intensities, especially in areas with long time series and high echo intensity. Furthermore, through ablation experiments, this paper proves that the temporal discriminator, spatial discriminator, and InceptionGRU all effectively improve the model’s prediction capability for radar echo images.

This study utilizes a new approach for short-term stock correlation forecasting using a combination of convolutional neural networks (CNN), bi-directional long and short-term memory (BiLSTM), and attention mechanisms to address the issue of information loss due to excessively long input time series data. The objective is to improve the accuracy of short-term equity correlation forecasts, which is essential for the efficient management of financial asset portfolios. The existing time series models usually only take into account the effect of stock historical data characteristics on stock correlation but ignore the fact that stock correlation is affected by multiple factors and undergoes an unknown unstable trend. More efficient methods or algorithms need to be devised to deal with financial data. This paper explores stock metrics and introduces a multi-factor model to quantify the factors affecting stock returns in order to calculate the correlation between stocks. Meanwhile, to cope with the problem of incomplete input data, this paper simultaneously incorporates stock return data that is directly related to the correlation data. In the CNN-BiLSTM-Attention (CLATT) model, the role of the CNN module is to obtain high-dimensional features from the input time series data, and the role of the BiLSTM module is to process the time series information to capture the time-series features in the stock return data and correlation data in the sample. In order to acquire the inherent correlation between the input stock return data and correlation data and to enhance the prediction accuracy of the model, the attention mechanism is finally implemented and used to optimize the weights of the BiLSTM output. In this paper, mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE) metrics, mean absolute percentage of error (MAPE), and the coefficient of determination $R^{2}$ , as well as four benchmark models, are used to evaluate the performance of the proposed model. The result shows that the CLATT model has 57.32% and 33% improvement in the correlation prediction accuracy of different stock portfolios compared to the single LSTM model, and the model prediction accuracy has been improved. Compared with the benchmark CNN-BiLSTM, it also has 56.06% and 28.87% improvement.

As an essential metal, copper has the advantages of electrical conductivity and ductility, which is widely used in power transmission, electronics manufacturing and machining. The fluctuation of copper price has a great impact on the industry, especially on the development of the national economy, so predicting copper price has a great significance for economic development. However, traditional time series prediction models’ prediction accuracy is low. Therefore, this paper proposes a Shanghai copper price forecasting model based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Convolutional Neural Network (CNN), Self Attention Gated Unit (SAGU), named CEEMDAN-CNN-SAGU. CEEMDAN decomposes and reconstructs the Shanghai copper price data. In this paper, the zero-cross rate of the Intrinsic Mode Function (IMF) components is calculated, the IMF component with large noise is removed, and the remaining IMFs and Residual term (Res) are reorganized to obtain the high-frequency and low-frequency components. CNN performs convolution operations on the reconstructed components to extract time series features. SAGU is a new time series data prediction model proposed in this paper. SAGU includes two gated units (a forgetting gate and an input gate), and two data processing modules (Self Attention (SA) and Transition (Tra)). The SA module is responsible for processing the input data, redistributing the weight coefficients of different data, and improving the model’s attention to important information. The Tra module improves the low limit of the forgetting gate’s output so that the gated unit can keep the cell state from the previous moment to the current moment as much as possible. This paper uses Shanghai copper trading data from Shanghai Futures Exchange as experimental data. The comparison experiment with the other eleven prediction models shows that the CEEMDAN-CNN-SAGU model outperforms other models in all evaluation indexes.

Video forgery detection is a critical aspect of digital forensics, addressing the challenges posed by the manipulation of video content. This paper presents a novel approach for video forgery detection using Deep Convolutional Neural Networks (DCNN). Leveraging the power of deep learning, our method aims to improve the accuracy and efficiency of object-based forgery detection in advanced video sequences. In the proposed approach, we build upon the foundation of an existing method, which utilizes Convolutional Neural Networks, and introduce innovative modifications to the DCNN architecture. These modifications include data preprocessing, network architecture, and training strategies that enhance the model’s ability to detect tampered objects in video frames. We conduct experiments on the SYSU-OBJFORG dataset, the largest object-based forged video dataset to date, with advanced video encoding standards. Our DCNN-based approach is compared with the existing method, demonstrating superior performance. The results show increased accuracy and robustness in detecting object-based video forgery. This paper not only contributes to the field of video forgery detection but also underscores the potential of deep learning, particularly DCNN, in addressing the evolving challenges of digital video manipulation. The findings open avenues for future research in the localization of forged regions and the application of DCNN in lower bitrate or lower resolution video sequences.

A crucial component within any structural health monitoring system is a pipeline leak detection mechanism, vital for preventing avoidable water loss. Contemporary literature employs machine learning and deep learning for detecting pipeline leaks and cross-correlation for leak localization. The major drawbacks in the existing methodologies are that machine learning and deep learning methods need two different architectures for leak detection and localization, and the cross-correlation needs two sensors with a denoising technique. The primary objective of this paper is to deploy a unified architecture capable of executing both the detection and localization of a leak without any denoising technique and with a single sensor. The proposed technique utilizes the data collected using an Acousto-optic sensor with two different pressures. This paper proposes a novel custom one-dimensional time-series DenseNet for leak detection and localization. The proposed method gives better accuracies compared with the existing one-dimensional DenseNet-121, three different one-dimensional convolutional neural networks (1DCNN), and cross-correlation for two different pressure datasets. The proposed method’s processing time is thirteen times less than the existing one-dimensional DenseNet-121, with the observed average leak detection and localization accuracy of 99.08%. The results state that the proposed novel custom one-dimensional time-series DenseNet accurately detects and localizes the leak with less time.

Currently, in the health area, a large amount of information is generated daily, enabling the creation of tools using the concepts of Machine Learning to help professionals make clinical decisions. In addition, due to the current scenario that we experienced in the Covid-19 pandemic, the development of tools that can assist in the diagnosis of this new disease has become essential. Besides several papers had explored deep learning algorithms for pneumonia classification, a recent survey showed they present some pitfalls and can not be used in real scenarios (eg. data sets from unreliable sources, training with small and imbalanced data sets, duplicated images due to improper merging data, problems with demographic differences among the patients such as age group, improper evaluation metrics, no use of external datasets for validation, no model interpretability). Moreover, the papers do not present a complete system, that can be tested in real scenarios. We aim to deal with these limitations and propose a framework to overcome the pointed pitfalls. We conducted a comprehensive analysis that underscores the significance of such an approach. Our efforts encompassed mitigating dataset biases and testing many popular Convolutional Neural Network models under a comprehensive evaluation. The inclusion of an external dataset fortified the credibility of our assessment. We engineered a prototype web platform with a Containerized architecture. Moreover, due to the overparameterization and black-box nature of deep learning models, it is difficult to understand the prediction results. We also explored tools to understand how the models make decisions. Through the experiments carried out in the ternary classification VGG16 network reached 89.7% of accuracy in the external datasets. In addition, the efficiency of these models in detecting the presence of diseases in patients, measured using the recall metric, was 0.96 for Covid-19 and 0.86 for Pneumonia, this result is of great importance since in the health area there is a great focus on avoiding false negatives.

This paper presents a hybrid algorithm for real-time instance segmentation of packages from scenes represented by 2D distance maps (range images). The paper introduces a novel approach combining deep learning-based methods and digital signal processing methods to enable accurate package recognition, using a small training dataset with high variability and distance measurement errors characteristic of Time-of-Flight-based scanning. Two convolutional neural networks with architecture optimized for training with a limited number of samples perform an initial segmentation of package components (sides and edges). An algorithm based on digital signal processing methods performs refinement of intermediate results, and combines package components into packages. Training and evaluation of the algorithm were performed on a custom dataset containing scenes of packages, shipping bags, and packaging of irregular shapes with various sizes, orientations, and degrees of occlusion, organized either in ordered stacks or arbitrary order. The convolutional neural networks provide a reliable distinction between components of packages and components of other types of packaging and surroundings. Package sides containing a sufficient number of distance points are correctly combined into packages. Thus, the proposed algorithm represents a solid basis for fully automated loading/unloading of packages with arbitrary sizes and materials from transport trailers and storage spaces. The dataset and annotations for box side surfaces are available at: https://dipteam.feit.ukim.edu.mk/results-package-detection.html.

The scintillator crystal is a crystalline material that can light under the influence of high-energy rays and be widely used for detecting high-energy particles. With the development of industrial CT and other fields, the demand for inorganic scintillator crystals with high melting points and high performance is increasing yearly, which results in the abandonment of the traditional laboratory mode of manufacturing inorganic scintillator crystals. For achieving the needs of high-standard inorganic scintillator crystals, the market attaches great importance to the intelligent manufacturing of scintillator crystals. One of the critical technical difficulties in intelligent crystal manufacturing is the numerical assessment of crystal quality. The traditional laboratory research and development mode relies on manually measuring crystal quality, which has many defects, such as the inability to be numerical, and the quality evaluation varies from person to person. Therefore, this paper proposes a methodology for assessing crystal quality based on the residual depth network. At the same time, in order to make crystal quality evaluation more refined, this paper uses a target detection method based on depth learning to determine the crystal quality assessment area. The target detection model in this article is developed based on the yolov5 deep learning framework, which is suitable for crystal target detection, and its detection accuracy is 98%. The residual depth network proposed in this paper is based on the network developed by ResNet-18, which is more suitable for crystal grading, and its accuracy of crystal quality evaluation reaches 84.8%. The quality evaluation of inorganic scintillation crystals is a solution to the problem of numerical representation of the quality of inorganic scintillation crystals, which opens up a technical channel for closed-loop control of intelligent crystal manufacturing processes.

The recommender system which gets higher in practical use in applying the Apriori algorithm in the early 2000s has revolutionized our daily life as it currently is widely used by big-tech platform companies. In the early stages of the development of recommender systems, services that can be provided to users were simply derived to the extent that only related products were recommended. However, the new research wave like deep learning-based recommender systems due to the development of information technology and the complexity of users’ online behavior extensively grabs researchers’ and academia’s attention in the field of recommender systems. This paper describes the algorithms and characteristics of the recent popular deep learning-based representative models such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), generative adversarial networks (GANs), and graph neural networks (GNNs) in the view of the sequential recommendation. The sequential recommendation of understanding user preferences in chronological order is useful for analyzing user-item interaction more accurately and flexibly. Therefore, the models specialized in sequential recommendation take advantage of understanding user behavior through temporal factors and improving recommendation quality by easily realizing the correlation between user and items. Also, the transformer-based model was developed to improve the problem of long-term dependency between users and items through factors, such as points, lines, and nodes, experienced in the early models of RNN and CNN and self-supervised learning (SSL)-based models, which are originally purposed to solve the data sparsity issues of recommender systems, will be discussed in this paper.

The growth of vision transformer (ViT) methods have been quite enormous since its features provide efficient outcome in image classification, and identification. Inspired of this beneficial, this paper propose an Enhanced vision transformer architecture (EViTA) model for pest identification, segmentation, and classification. The as of late found that, compare to machine learning, Convolutional neural network algorithms the ViT has providing trusted results on image classification. Motivated by this, in this paper, we concentrate on the best way to learn dual barch segment representations in ViT models for image arrangement. Based upon its features, here propose a double layer transformer encoder to integrate pest image segments of various sizes of pest images to create more grounded image highlights. The current study uses, three pest datasets that affects peanut crops such as Aphids (IP102 Dataset), Wireworm (IP102 Dataset), and Gram Caterpillar collected from public available repository. Our methodology processes small segment and huge segment of tokens with two separate parts of various computational intricacy also, these tokens are then combined simply by consideration numerous times to complete one another. The taken datasets’ are preprocessed utilizing the characteristic by using moth flame optimization (MFO), and flatten the images by using linear projector methodology to enhance the missing quality in pest images, and afterward normalization methods are executed to switch it over completely in to mathematical arrangement. This processed information is standardized further utilizing the self attention in StandardScaler procedures are carried out for choosing the ideal highlights in the dataset accordingly having huge effect towards affecting pest image predictions. These ideal highlights are at last taken care of into the EViTA model and the outcomes created are considered in contrast to the cutting edge models which at last legitimize the predominance of the proposed EViTA+PCA+MFO model in pest image prediction with high accuracy rate. Broad trials show that our methodology performs better compared to or on standard with a few simultaneous deals with vision transformer, notwithstanding productive CNN models.

Unmanned Aerial Vehicles and the increasing variety of their applications are raising in popularity. The growing number of UAVs, emphasizes the significance of drones’ reliability and robustness. Thus, there is a need for an efficient self-observing sensing mechanism to detect real-time anomalies in drone behavior. Previous works suggested prediction models from control theory, yet, they are complex by nature and hard to implement, while Deep Learning solutions are of great utility. In this paper, we propose a real-time framework to detect anomalies in drones by analyzing the sound emitted from them. For this purpose, we construct a hybrid Deep Learning based Transformer and a Convolutional Neural Network inspired by the well-known VGG architecture. Our approach is examined over a dataset that is collected from a single microphone set located on a micro drone in real-time. Our approach achieves an F1-score of 88.4% in detecting anomalies and outperforms the VGG-16 architecture. Moreover, the framework presented in this paper reduces the number of parameters of the well-known VGG-16 from 138M, into a shrunk version with 3.6M parameters only. Additionally, our real-time approach, results in a smaller number of parameters in the neural network, and yet yields high accuracy in anomaly detection in drones with an average inference time of 0.2 seconds per second. Moreover, with an earphone that weighs less than 100 grams on top of the UAV, our method is shown to be beneficial, even in extreme conditions such as a micro-size dataset that is composed of three hours of flight recordings. The presented self-observing method can be implemented by simply adding a microphone to drones and transmitting the captured audio for analysis to the remote control or performing it onboard the drone using a dedicated microcontroller.

As an interdisciplinary course, Machine Vision combines AI and digital image processing methods. This paper develops a comprehensive experiment on forest wildfire detection that organically integrates digital image processing, machine learning and deep learning technologies. Although the research on wildfire detection has made great progress, many experiments are not suitable for students to operate. Also, the detection with high accuracy is still a big challenge. In this paper, we divide the task of forest wildfire detection into two modules, which are wildfire image classification and wildfire region detection. We propose a novel wildfire image classification algorithm based on Reduce-VGGnet, and a wildfire region detection algorithm based on the optimized CNN with the combination of spatial and temporal features. The experimental results show that the proposed Reduce-VGGNet model can reach 91.20% in accuracy, and the optimized CNN model with the combination of spatial and temporal features can reach 97.35% in accuracy. Our framework is a novel way to combine research and teaching. It can achieve good detection performance and can be used as a comprehensive experiment for Machine Vision course, which can provide the support for talent cultivation in machine vision area.

Computer vision research in detecting and classifying the subtype Acute Lymphoblastic Leukemia (ALL) has contributed to computer-aided diagnosis with improved accuracy. Another contribution is to serve as an assistant and second opinion for doctors and hematologists in diagnosing the ALL subtype. Early detection can also rely on computer-aided diagnosis to determine initial treatment. The purpose of this study is to review the progress of research in the detection and classification of ALL subtypes. The method’s discussion focuses on the application of deep learning to the domain of object detection and classification. Motivations, challenges, and future research recommendations are thoroughly discussed to improve understanding and progress in this field of study. The study was carried out methodically by analyzing a collection of papers on the detection and classification of ALL subtypes published in science direct, IEEE, and PubMed from 2018 to 2022. The analysis of this paper field is included in the results of the selected paper. The paper selection from among 65 papers was based on inclusion and exclusion methods. Based on research methods and objectives, papers are divided into two large groups. The first group discusses the classification of ALL subtypes, while the second group discusses the detection of ALL subtypes. The discussion of prior research reveals some challenging issues and future work, such as the limited availability of the ALL subtypes dataset, the high computational complexity of the deep learning model, and further exploration of transformers in computer vision as a reference for research gaps that can contribute to future research.

Within Artificial Intelligence, Deep Learning (DL) represents a paradigm that has been showing unprecedented performance in image and audio processing by supporting or even replacing humans in defect and anomaly detection. The railway sector is expected to benefit from DL applications, especially in predictive maintenance applications, where smart audio and video sensors can be leveraged yet kept distinct from safety-critical functions. Such separation is crucial, as it allows for improving system dependability with no impact on its safety certification. This is further supported by the development of DL in other transportation domains, such as automotive and avionics, opening for knowledge transfer opportunities and highlighting the potential of such a paradigm in railways. In order to summarize the recent state-of-the-art while inquiring about future opportunities, this paper reviews DL approaches for the analysis of data generated by acoustic and visual sensors in railway maintenance applications that have been published until August 31st, 2021. In this paper, the current state of the research is investigated and evaluated using a structured and systematic method, in order to highlight promising approaches and successful applications, as well as to identify available datasets, current limitations, open issues, challenges, and recommendations about future research directions.

Detecting objects in aerial images is a challenging task due to the large-scale variations and arbitrary orientations with tiny instances. A new multi-scale transformer-based aerial objects detector called MStrans is proposed in this paper to deal with the challenges in aerial detection. To detect remote instances, MStrans adopts a multi-scale patch embedding transformer (MViT) to extract the global features of the image effectively. Furthermore, to capture the different discriminant features for classification and regression branch tasks, the partial interactive fusion module (PIFM) is designed to enhance the semantic expression of the key features of classification and regression tasks by using the strategy of interactive modeling of adjacent layer features. In addition, considering that the transformer may worsen the local feature details while capturing long-distance feature dependencies, this paper designs a global to local interactive fusion module (GLIFM). It uses the advantage of convolution to extract local features to enrich the detailed information in the transformer. Experiments were carried out on DOTA and DIOR datasets, and the MStrans achieves superior detection performances compared with other approaches.

Air Quality Index (AQI) is the crucial foundation for measuring air quality, which reflects the influence of air quality on people’s health and life to a certain extent. In this paper, a hybrid AQI prediction model based on Convolutional Neural Network (CNN) and Attention Gate Unit (AGU) is proposed to deal with the problems of the “vanishing gradient” and “exploding gradient” of Recurrent Neural Network (RNN). AGU is a new model proposed in this paper to embed the attention mechanism and Data Adjustment Module (DAM) into the gated unit. The attention mechanism enhances the learning ability of the gated unit, and the DAM makes the gated unit more sensitive to historical data learning. In this model, CNN plays a role in extracting features from time series data. AGU can make differentiated learning of historical data and finally produce prediction results. The model evaluation indexes used in the experiments are Mean Absolute Error (MAE), Mean Square Error (MSE), and R Squared (R2). The experimental results show that the overall performance of the AQI prediction model based on CNN-AGU is superior to that of other models by comparing with the other nine models on the same data set.

Dataflow-scheduling techniques for convolutional neural networks (CNNs) are extensively studied to minimize the off-chip memory access. However, the efficiencies of the previously proposed techniques are limited because their optimizations only consider the general hardware such as FPGA and GPU. To overcome this limitation, this paper proposes dataflow scheduling for vector-SIMD DSP to minimize the energy consumption for the off-chip memory access. First, the proposed technique attempts to group as many given layers as possible. For grouping the layers, the tiles in different layers are executed in sequence without the off-chip memory access except the first and the last layers in the group. The length of the grouped layers is determined with regard to the minimization of the energy consumption of off-chip memory by estimating the proposed energy model of the off-chip memory. However, grouping the layers results in the additional computation. To minimize this overhead, this paper solves the optimization problem for in the grouped layers. Second, for layers that cannot be grouped, the tiling along the W-axis is not considered, to maximize the size of the overlapped data in consecutive tiles. Consequently, the reuse of the overlapped data in the on-chip buffer is maximized, thereby reducing the energy consumption by the off-chip memory. For evaluation, a cycle-accurate simulation environment is established to measure the energy consumption of the off-chip memory by tracing the data between a vector-SIMD DSP and an off-chip memory. The experimental results show that compared with the baseline tiling and scheduling techniques, the proposed technique reduces the energy consumption by an average of 51% for CNN applications such as Tiny YOLOv2, MobileNetv1, VDSR.

When a learned model has high accuracy under familiar settings (internal testing) and a big drop in accuracy under slightly different circumstances (external testing) we suspect it is using shortcuts to make decisions. This problem is known as shortcut learning. In medical imaging, shortcuts are undesired and unintended features that the model relies on to perform diagnosis. Shortcut-based solutions using medical images could lead to false diagnoses and have dangerous implications for patients. In the current COVID-19 era, a large set of papers have been published proposing the use of deep convolutional neural networks to perform diagnosis or triage of COVID-19 from chest X-rays (CXRs). These studies are reporting high accuracies which could be misleading and overestimated. To our knowledge, none of the currently published papers with high performance reported testing on samples from truly unseen data sources. Studies which did, have noticed a significant performance drop when testing on unseen sources indicating a failure to generalize. In this paper, we elucidate the generalization challenge of deep learning based models trained for disease diagnosis. We use the example of COVID-19 diagnosis from CXRs. Solutions that mitigate shortcut learning are introduced and experimentally shown to be effective. Our proposed methods enable the models to have a statistically significantly reduced performance drop-off on unseen data sources. Thus, lowering the performance drop to only 9% instead of 20%. The issues with convolutional neural networks addressed here generally apply to other imaging modalities and recognition problems, as shown.

Image transmission over Low-Power Wide Area Networks (LP-WAN) protocols has always been a difficult task since it necessitates high data rates and high energy consumption. Long Range (LoRa) is one such protocol, which is excellent for transferring data over long distances but has generated severe doubts regarding the viability of image transmission due to its low data rate. This paper demonstrates the application results of an integrated LoRa and Deep Learning-based computer vision system that can efficiently identify grape leaf diseases using low-resolution images. In particular, the focus in this paper is to combine the two technologies, LoRa and Deep Learning, to make the transmission of the images and the identification of the diseases possible. To achieve this objective, the framework utilizes a combination of on-site and simulation experiments along with different LoRa parameters and Convolutional Neural Model (CNN) model fine-tuning. Based on the evaluation, the proposed framework proved that the transmission of images using LoRa is possible within the protocol limitations (such as limited bandwidth and low duty cycle). Our fine-tuned model can efficiently identify grape leaves diseases. The technique is both efficient and adaptive to the specifics of each leaf disease, while it does not need any training data to adjust parameters. It is worth noting that today, end-user trust in Machine and Deep Learning models has increased significantly because of novel solutions in the field of Explainable Artificial Intelligence (XAI). In this study, we use the Grad-CAM method to visualize the output layer judgments of the CNN. The disease’s spot region is highly activated, according to the visualization findings. This is how the network distinguishes between different grape leaf diseases.

It is generally noticed that increasing the number of convolutional layers in generic image classification procedures proves to be detrimental to model performance in terms of validation accuracy and loss. Apart from vanilla CNNs, we have state-of-the-art (SOTA) architectures such as ResNet50 (and its variants) which show that through the use of skip-connections, higher performance metrics are attainable through deeper architectures. However, most evaluative metrics converge on a log scale as we go deeper with diminishing gradient of the metrics’ curves. Given these two contrasting speculations, in this paper, we implement various vanilla and SOTA CNNs for the diagnosis of one of the most common forms of breast cancer - invasive ductal carcinoma (IDC) - to examine and understand the feasibility of implementation of SOTA CNNs through transferred weights when juxtaposed with vanilla CNNs (and LeNet-5) of varying configurations in terms of their performance metrics and other parameters. In this paper, we solve the dual-objective of studying behavioural aspects of avant-garde CNN models (more specifically, VGG16, VGG19, ResNet50, ResNet50V2, MobileNetV2, and DenseNet121) and proper diagnosis of IDC through intermediate neural activations to critically evaluate and theorize the performance of different models. We notice that among all the models, only VGG16, VGG19, LeNet-5 and a selected vanilla CNN through an optimization procedure were the ones to attain the best metrics, shared amongst them.

Chest X-ray is a prevalent medical imaging modality for detecting lung diseases. The clinical analysis of X-ray images is usually conducted by radiologists, who represent valuable human resources. In practice, situations with insufficient radiologists to timely analyze large quantities of X-ray data are very common. Accordingly, developing an automated computer-aided lung disease classification system is beneficial to facilitate diagnoses. However, due to restrictions of costs and time, collecting large amounts of accurately labeled X-ray images to train a machine learning based diagnosis system is challenging. Another limitation is the class imbalances present in datasets. Facing these challenges, we investigate the effectiveness of using generative models, particularly generative adversarial networks (GANs), to synthesize new data to tackle the issue of data paucity and class imbalances. To this end, it should be noted that few existing works have studied the effect of generated image quality on the performance of different learning models, particularly in medical imaging. Therefore, the current paper represents one of the first comprehensive investigations into the impact of synthetic image generation on classifier performance, which is empirically elucidated by a comparative analysis of a simpler deep convolutional GAN to a more complex progressive GAN design. Another contribution of this paper is a multi-scale convolutional neural network (CNN) architecture, which can take advantage of image features at different scales for better learning from scratch. Altogether, to verify the robustness of using GANs to augment datasets, we compare various data augmentation approaches, when applied to different network architectures, including transfer learning, learning-from-scratch CNNs, state-of-the-art ResNet, EfficientNet, DenseNet, and the proposed multi-scale CNN. Specifically, testing on two publicly available datasets, the obtained results show that using finer images synthesized from GANs with the proposed multi-scale CNN achieved good classification performance, under a wide range of operating conditions.

The Exchange rate affects the economic development of various countries. To grasp the information of the changeable exchange rate in time, it is necessary to predict the exchange rate price. This paper proposes the CNN-TLSTM model to predict the United States Dollar/Chinese yuan (USD/CNY) exchange rate closing price of the next trading day. The model consists of two parts, namely convolutional neural networks(CNN)and tanh long short-term memory (TLSTM). The function of CNN is to extract feature factors from the input data. TLSTM is used to receive the output data of CNN for prediction, and finally obtain the prediction result. TLSTM is a new model proposed in this paper to improve the internal structure of the long short-term memory (LSTM). Its advantage is to change the range of the output value of the input gate, retain more data features, and prevent the output value of the input gate in LSTM from being overfitting. This paper selects the exchange rate of USD/CNY data and some stock data for each trading day from January 2, 2006, to October 30, 2020, as the experimental data. To prove the effectiveness of the CNN-TLSTM prediction model, the model is compared with multilayer perceptron (MLP), CNN, recurrent neural network (RNN), LSTM, and CNN-LSTM models. mean absolute percentage error (MAPE), mean square error (MSE), and R-squared (R2) are used for comparative analysis. The experimental results show that the CNN-TLSTM model has the best predictive effect on the USD/CNY exchange rate closing price of the next trading day.

Music is a heavenly way of expressing feelings about the world. The language of music has vast diversity. For centuries, people have indulged in debates to stratisfy between Western and Indian Classical Music. But through this paper, an understanding can be fabricated while differentiating the types of Indian Classical Music. Classical music is one of the essential characteristics of Indian Cultural Heritage. Indian Classical Music is divided into two major parts, i.e. Hindustani and Carnatic. Models have been sculptured and trained to classify between Hindustani and Carnatic Music. In this paper, two approaches are used to implement classification models. MFCCs are used as features and implemented models like DNN (1 Layer, 2 Layers, 3 Layers), CNN (1 Layer, 2 Layers, 3 Layers), RNN-LSTM, SVM (Sigmoid, Polynomial & Gaussian Kernel) as one approach. A 3 channels input is created by merging features like MFCC, Spectrogram and Scalogram and implemented models like VGG-16, CNN (1 Layer, 2 Layers, 3 Layers), ResNet-50 as another approach. 3 Layered CNN and RNN-LSTM model performed best among all the approaches.

High-resolution temperature forecasting can often prove to be challenging for conventional machine learning models as temperature is highly seasonal and varies with the time of the year as well as with passing hours of the day. In most cases, only the daily extremes or mean temperatures are provided by temperature forecasting methods. However, with the growing availability of data and the development of deep neural networks (DNNs) capable of detecting complex relationships, high-resolution temperature forecasting is becoming easier. Typically, historical temperature data along with multiple meteorological sensor data is used for temperature forecasting which increases the complexity of the system making it harder and costlier to implement physically. In this paper, high-resolution hourly temperature forecasting is performed using only historical temperature data. The paper presents a comparative analysis among four popular DNNs-simple recurrent neural network (SRN), gated recurrent unit (GRU), long-short term memory (LSTM), convolutional neural network (CNN), and two hybrid models-CNN-LSTM parallel network and GRU-LSTM parallel network trained on Beijing temperature dataset. Experimental results showed GRU-LSTM parallel network obtained the lowest RMSE (1.691° C) whereas CNN has the best computational efficiency obtaining a slightly worse RMSE (1.759° C). Additionally, a robustness analysis is performed on temperature data from four additional geographically diverse locations (Toronto, Las Vegas, Seattle, and Dallas) which reveals GRU to be the most consistent algorithm. Finally, the paper establishes a correlation between the model performance and the dataset based on their variance and mean absolute deviation with reference to the training dataset.

The safety of workers at construction sites is one of the most important aspects that should be considered while performing their required tasks. Many rules and regulations have been implemented in the UAE to reduce injuries and fatalities in the jobsites. However, the number of accidents continues to increase. For instance, an accident category of fall-from-height is considered as the top cause of injuries and fatalities. Thus, this paper develops a novel technique that monitors the workers whether they are complying with a safety standard of the Personal Fall Arrest System (PFAS). This paper establishes a real time detection algorithm based on a Convolutional Neural Network (CNN) model in order to detect two main components of the PFAS that are, safety harness and life-line, in addition to a standard safety measure of using a safety helmet. The YOLOv3 algorithm is adopted for a deep learning network used to train the desired model. The model achieved an accuracy rate of 91.26% and around 99% precision. Moreover, the overall recall of the model was 90.2%. The obtained results verify the effectiveness of our proposed model in construction sites to control potential violations and to avoid unnecessary accidents. The main contribution of this paper is to provide an AI-based image detection framework to mitigate the likelihood of fall-from-height accidents.

Prescription dispensing accuracy is of paramount importance for all hospitals. However, human errors are inevitable due to multiple reasons, such as fatigue, stress, heavy workload, lack of effective verification measures, mismanagement. Such human errors pose serious safety and health concerns on the part of patients and may as well lead to a series of medical disputes. Based on induced deep learning, this paper proposes a real-time Blister Package Identification System (BPIS) to assist pharmacists’ drug verification and dispensing. Under the guidance of the induction strategy, image preprocessing is introduced to form a standardized image containing the front and back side of the blister package, which is subsequently sent to CNN-based object identification network for feature extraction and identification. This preprocessing method allows the identification system to promote the deep learning system to focus on feature learning to obtain more information about the appearance of the package ruling out confounding factors such as background noise, size, shape or positioning. In addition, this article collects and establishes an image dataset of adult lozenges. Under this dataset, this paper verifies the enhancement of Induced Deep Learning (IDL) on YOLO v2, ResNet, and SENet. By optimizing the deep learning identification network with the help of the embedded technology and a two-side extraction mechanism, a real-time BPIS is built. Long-term tests in hospitals prove the effectiveness of the proposed system.

In recent years, with the rapid development of Internet technology, online shopping has become a mainstream way for users to purchase and consume. Sentiment analysis of a large number of user reviews on e-commerce platforms can effectively improve user satisfaction. This paper proposes a new sentiment analysis model-SLCABG, which is based on the sentiment lexicon and combines Convolutional Neural Network (CNN) and attention-based Bidirectional Gated Recurrent Unit (BiGRU). In terms of methods, the SLCABG model combines the advantages of sentiment lexicon and deep learning technology, and overcomes the shortcomings of existing sentiment analysis model of product reviews. The SLCABG model combines the advantages of the sentiment lexicon and deep learning techniques. First, the sentiment lexicon is used to enhance the sentiment features in the reviews. Then the CNN and the Gated Recurrent Unit (GRU) network are used to extract the main sentiment features and context features in the reviews and use the attention mechanism to weight. And finally classify the weighted sentiment features. In terms of data, this paper crawls and cleans the real book evaluation of dangdang.com, a famous Chinese e-commerce website, for training and testing, all of which are based on Chinese. The scale of the data has reached 100000 orders of magnitude, which can be widely used in the field of Chinese sentiment analysis. The experimental results show that the model can effectively improve the performance of text sentiment analysis.

With the popularity of the internet, the expression of emotions and methods of communication are becoming increasingly abundant, and most of these emotions are transmitted in text form. Text sentiment classification research mainly includes three methods based on sentiment dictionaries, machine learning and deep learning. In recent years, many deep learning-based works have used TextCNN (text convolution neural network) to extract text semantic information for text sentiment analysis. However, TextCNN only considers the length of the sentence when extracting semantic information. It ignores the semantic features between word vectors and only considers the maximum feature value of the feature image in the pooling layer without considering other information. Therefore, in this paper, we propose a convolutional neural network based on multiple convolutions and pooling for text sentiment classification (variable convolution and pooling convolution neural network, VCPCNN). There are three contributions in this paper. First, a multiconvolution and pooling neural network is proposed for the TextCNN network structure. Second, four convolution operations are introduced in the word embedding dimension or direction, which are helpful for mining the local features on the semantic dimensions of word vectors. Finally, average pooling is introduced in the pooling layer, which is beneficial for saving the important feature information of the extracted features. The verification test was carried out on four emotional datasets, including English emotional polarity, Chinese emotional polarity, Chinese subjective and objective emotion and Chinese multicategory. Our approach is effective in that its result was up to 1.97% higher than that of the TextCNN network.

Visual attention mechanism has been widely used in computer vision and plays a vital role in the research of human action recognition. In this paper, we explore a novel moving target detection mechanism for human action recognition and propose a new 3D CNN (3D Convolutional Neural Network) model, dubbed LaM-2SRN (Local Features Enhanced and Moving target detected 2Stream-ResNet) for extracting and learning attention-enhanced spatiotemporal features. The contributions of this paper are as follow: First, the traditional CAM (Class Activation Map) based visual attention algorithm is used to obtain the optical flow information of the human region, thus eliminating the influence of irrelevant optical flow information (such as background clutter). Second, the ViBe algorithm is used to identify the moving target in the continuous frame, retain the optical flow information of the moving target, and make it complement with the optical flow information of the human region, to obtain a complete motion descriptor. After the motion information is marked in the video frame, we put the marked video frames and the original video frames into the 3D CNN and 2D CNN models respectively, to acquired mixed descriptor, which serves as the basis for video classification. Different from most of the previous target detection algorithms, this paper not only detects the human target, but also detects the moving foreground targets in the video, so that the pre-trained CNN model can obtain more complete motion information from videos. We use RGB-only video data for evaluation on two benchmarks: UCF101 and HMDB51, the experimental results demonstrate that our LaM-2SRN is comparable to the previous state-of-the-art algorithms.

The emerging of deep neural networks, especially the convolutional neural network (CNN), substantially promotes the fast development of brainware processors in object detection. However, the vast network architecture brings severe challenges to the design of brainware processor, which requires a large number of logic gates and memories. Therefore, a compact brainware processor with less memory and logic gate has a high demand in object detection. Typically, the object detection involves single-shot and multi-shot detectors in accordance with different detection principle. In the early stage, the multi-shot detector has a leading role in solving object detection issues, such as region-based convolutional neural networks (R-CNNs), faster R-CNNs etc. However, the multi-shot detector suffers from a low detection rate comparing with the single-shot detector. The you only look once (YOLO) algorithm, as the state-of-the-art real-time object detection algorithm, receives extensive attention from the academics and industry. Particularly, the lightweight YOLO algorithm, tiny YOLO3, has excellent potential for circuit design of compact brainware processor. Nonetheless, systematic studies of tiny YOLO3 are still missing up to the present. This paper offers a thorough review of the tiny YOLO3 algorithm, which can fill the gap in the field of object detection. Furthermore, the open solutions of compressing the tiny YOLO3 algorithm are proposed from the aspects of algorithm, hardware and emerging technology. The comprehensive study presented in this paper can not only enhance understanding of the tiny YOLO3 algorithm for researchers or engineers but also make a significant contribution to accelerating the development of compact brainware processor.

Using the optical camera in remote sensing is limited in various environmental conditions. This paper presents a system of combining deep learning and image transform algorithms to detect landslide location in satellite images. In the deep learning part, a convolution neural network is used to classify satellite images contain landslides. From landslide images classified, in order to accurately identify landslides under different lighting conditions, this paper proposes a transformation algorithm Hue - Bi-dimensional empirical mode decomposition (H-BEMD) to determine the landslide region and size. After the location of landslide is detected, we discover the size change of the landslide based on different time points. In this study, we record an accuracy of up to 96% in the classification process, and the accuracy of landslide location almost absolute.

Gold immunochromatographic assay (GICA) is a widespread rapid detection method with less cost but high efficiency. It is easy to operate and dispense with professional staff and equipment, which conforms to the trend of point-of-care testing that advocated by modern medicine. With the development and progression of medical detection technology, the qualitative analysis that could be easily performed with the naked eye is not satisfying anymore. In recent years, improving the performance of quantitative analysis of the GICA has become a hot research topic. However, the GICA is susceptible to noise interference due to various factors when used in the qualitative analysis in clinics. The rise of artificial intelligence has provided us with new ideas and directions. As a popular neural network in deep learning, convolutional neural network (CNN) has achieved excellent results in image processing and has been widely applied to many fields, including biomedical engineering. In this paper, CNN is applied to the image segmentation of gold immunochromatographic strip. The grayscale features of the pre-processed images are learned by the established CNN network, and then, the control and test lines are accurately extracted and further quantitative analysis is performed. The results show that the method proposed in this paper has a good segmentation effect on the GICA, and it also provides a new scheme for the quantitative analysis of the GICA.

Google Ads is an advertising agency that provides ads to advertisers. Advertisers match the user's search terms and push ads by selecting keywords related to their ad content. Keywords can determine the type of users an advertiser pushes, the effectiveness of the ad promotion, and the sales of the ad product. Automatically selecting keywords that are satisfactory to advertisers from a large number of keywords provided by Google Ads is the main task of this paper. But there is not too much time for the model to judge whether keywords are selected, choosing correct keywords in the shortest time is another task of this paper. Therefore, a structure of the model that can get some useful keywords for advertisers is designed and an improved multi-objective particle swarm optimization algorithm is proposed to achieve this multi-objective task. These are also the main contributions of this paper. To accomplish this multi-objective task, many technical issues need to be overcome, such as the mixed language problem, the imbalance problem, the problem of extracting features from corpora and so on. This paper proposes a corpus selection method to solve the mixed problem of Chinese and English in keywords, word embedding method to solve the representation of keywords, re-sampling to solve data imbalance problem, improved convolutional neural network (CNN) to solve classification problem, and a multi-objective particle swarm optimization algorithm (MOPSO) to achieve neural structure search of CNN so that the effect of the classification is improved and the training time is reduced. The keyword selection problem is solved with the combination of evolutionary computing, deep learning, machine learning, and text processing techniques. Experimental results show that the proposed algorithm greatly improved the accuracy of keyword selection and shortened the time of selecting keywords. Therefore, this algorithm has a good application value.

Image similarity learning aims to exploit the correlation between different images by learning image appropriate common features. In recent years, the previous CNN-based methods have directly learned the similarity between image features, which effectively improves the learning efficiency of image similarity. However, it has the following limitations: (1) The extracted image features are too single to describe the content of the image accurately; and (2) the network training is limited by the amount of dataset size. Data augmentation and multi-feature fusion have been demonstrated to appropriately improve the model generalization ability for various vision tasks. This paper integrates these methods into the network structure to design three multi-feature fusion network. The input of network training adopts the same data from the multi-input method to realize data augmentation, and the diversity of extracted image features is greatly improved by fusing different features. Then, the trained network of different dataset size is utilized to verify the network training adaptability of multi-feature fusion network. Moreover, the influence of loss function and optimization algorithm on the learning efficiency of complicated networks have been studied. The experimental results show that our proposed method has excellent performance on the self-collected XPU and Totally-Looks-Like (TLL) dataset, the learning and model generalization ability of multi-feature fusion network are significantly improved through data augmentation and multi-feature fusion. The multi-feature fusion network proposed in this paper has strong adaptability to network training.

The dilated convolution algorithm, which is widely used for image segmentation, is applied in the image classification field in this paper. In many traditional image classification algorithms, convolution neural network (CNN) plays an important role. However, the classical CNN has the problem of consuming too much computing resources. To solve this problem, first, this paper proposed a dilated CNN model which is built through replacing the convolution kernels of traditional CNN by the dilated convolution kernels, and then, the dilated CNN model is tested on the Mnist handwritten digital recognition data set. Second, to solve the detail loss problem in the dilated CNN model, the hybrid dilated CNN (HDC) is built by stacking dilated convolution kernels with different dilation rates, and then the HDC model is tested on the wide-band remote sensing image data set of earth's terrain. The results show that under the same environment, compared with the traditional CNN model, the dilated CNN model reduces the training time by 12.99% and improves the training accuracy by 2.86% averagely, compared with the dilated CNN model, the HDC model reduces the training time by 2.02% and improves the training and testing accuracy by 14.15% and 15.35% averagely. Therefore, the dilated CNN and HDC model proposed in this paper can significantly improve the image classification performance.

As the leading mobile operating system, Android powers critical infrastructure and personal devices across sectors such as finance and healthcare and a wide range of user scenarios. However, its open-source nature presents considerable security challenges. Exploiting vulnerabilities in native and custom permissions, malicious API calls, intents, signatures, and manifests, threat actors can gain access to sensitive data and device control. The continuously evolving landscape of Android malware necessitates robust and generalized detection methods. While traditional machine learning (ML) models have been employed to address this issue, they have limitations. Focusing on single datasets can impede both generalization and effective detection. Further, the dynamic nature of malware renders many traditional methods inadequate for providing comprehensive and real-time protection. This paper addresses this critical need by proposing DeepImageDroid, an advanced and efficient deep learning (DL) framework for Android malware detection. DeepImageDroid harnesses the combined power of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), utilizing three diverse Android malware datasets. This hybrid approach significantly improves detection accuracy and model generalization compared to existing solutions. By employing the weighted average ensemble method, DeepImageDroid achieved a remarkable 96% accuracy.

The rapid advancement of artificial intelligence (AI), particularly in text-to-image generative models, has led to a proliferation of synthetic images. This progress, while remarkable, raises concerns about misuse in fraudulent activities. To address this issue, we propose a Convolutional Neural Network (CNN)-based approach for classifying AI-generated images from multiple generators. We introduce a gated CNN model that leverages mixed datasets for improved training efficiency and performance. This approach eliminates the need for extensive tuning with each new dataset and mitigates the risk of catastrophic forgetting. Our experiments demonstrate that the gated CNN model slightly outperforms traditional single CNN models, providing a more robust solution for identifying AI-generated images. This paper presents a comprehensive comparison of methods and offers insights into enhancing the classification of AI-generated images.

In recent years, deep learning (DL) has become indispensable in medical image segmentation (MIS), proven by numerous studies showcasing its effectiveness. This paper presents two significant original contributions and conducts a comprehensive thematic evaluation of DL approaches in MIS. Diverging from traditional surveys that categorize DL literature into multiple groups and analyze individual works within each category, we adopt a multi-level classification approach. This method organizes existing literature from a broad overview to finer details. Furthermore, to address the inherent challenges of this field, we introduce an innovative method called L-UNet. The L-UNet method strategically utilizes filters with reduced parameters to construct a U-shaped convolutional neural network. This not only demonstrates impressive performance but also alleviates complexities associated with architectural parameters. Extensive evaluation across multiple datasets, including lung image segmentation, PH2 skin cancer segmentation, liver image segmentation, Chest X-ray segmentation, and COVID-19 chest X-ray segmentation, highlights the exceptional performance of L-UNet. Notably, experimental results showcase competitive accuracies, achieving figures such as 99.15% for lung image segmentation, 99.53% for liver image segmentation, 95.45% for PH2 skin cancer segmentation, and 98.99% for Chest X-ray segmentation. Moreover, for COVID-19 chest X-ray segmentation, a commendable accuracy of 93.45% is observed. The simplification of L-UNet renders it a highly practical choice for deployment in resource-limited environments or real-world scenarios necessitating faster performance and enhanced efficiency.

Deep learning has been widely used in medical image processing, which has sparked the development of a wide range of applications and led to a notable increase in the number of therapeutic and diagnostic options available for a range of medical imaging problems. In the era of the Internet of Things (IoT), safeguarding the security and privacy of medical data is crucial to the advancement of sophisticated diagnostic applications for medical imaging. Deep learning-based brain tumor detection in smart health care systems with privacy preservation is proposed in this paper. The system under consideration is organized into three discrete stages that are then combined to provide an all-encompassing blueprint. During the first phase, patients with brain tumors are the primary target of an efficient healthcare system that is introduced. A Microsoft-based operating system-compatible application has been developed to accomplish this. Patient data is secure and only available to the hospital and the individual patient, which enables patients to engage with the system both locally and virtually. To obtain the anticipated outcomes, the user must first submit the patient’s MRI scan and then enter a special 10-digit code. In the second part, the authors develop a deep learning-based tumor identification platform which also incorporates the AES-128 algorithms and PBKDF2 for secure medical image storage on the server and data transmission via the internet from the client to the server and back to the client upon prediction. The proposed approach integrates ResNet-50, Inception V3, and VGG-16 architecture to build a Convolutional Neural Network (CNN)-based brain tumor diagnosis system. These architectures are enhanced through significant pre-processing, SGD, RMSprop, and Adam optimization. Our research focuses on the application of cutting-edge methods to maintain confidentiality and accomplish precise tumor diagnosis, underscoring the importance of privacy preservation. Our micro-average findings were the best, with 99.92% accuracy, 99.99 % Area Under the Curve (AUC), 99.9 % precision, 99.92 % recall, and 99.92 % F1-score. Moreover, significant influence on tumor categorization was demonstrated when the experimental outcomes of the modified models were contrasted with multiple CNN-based architectures through the use of critical performance criteria.

In recent years, numerous studies have investigated the development of methods for video quality assessment (VQA). These studies have predominantly focused on specific types of video degradation tailored to the application of interest. However, natural videos or recent videos generated by users (UGC) present complex distortions that are not easy to model. Consequently, most current VQA approaches struggle to achieve high performance when applied to these videos. In this paper, we propose a novel Transformer-based architecture that extracts spatial distortion features and spatio-temporal features from videos in two specialized branches. The spatial distortion branch leverages a transfer learning strategy where a standard ViT is pre-trained using a masked autoencoder (MAE) self-supervised learning task, and then fine-tuned to predict the distortion type of corrupted images from the CSIQ database. The features from this branch capture degradation at the level of individual frames. On the other hand, the second branch employs a 3D Shifted Windows Transformer (Swin-T) to extract spatio-temporal features across multiple frames. Once again, we use transfer learning to extract rich features by pre-training this 3D Swin-T model on a video dataset for human action recognition. Finally, a temporal memory block hinged on an attention recurrent neural networks is proposed to predict the final video quality score from the spatio-temporal sequence of features. We evaluate the performance of our method on two popular UGC databases, namely KoNViD-1k and LIVE-VQC. Results show it outperforms state-of-the-art models on the KoNViD-1k database, achieving a SROCC performance of 0.927 and a PLCC of 0.925, while also delivering highly competitive results on the LIVE-VQC database.

The informativeness of data has always been of great interest within the machine learning community. Nowadays, with the skyrocketing advancement of artificial intelligence and massive volumes of noisy data, it becomes even more essential to develop robust and effective methods for training data optimization. Existing approaches are mostly based on empirical trial and error, with either stochastic or deterministic data reduction strategies. The key limitation of such solutions is that they do not consider the overall informativeness of the resulting training dataset. In this paper, a novel approach for quality-aware dataset optimization by initial assessment of its informativeness is proposed. As a metric of informativeness, entropy values are calculated over the target dataset. To alleviate the computational complexity, an initial clustering of the dataset is performed, and the entropy of each cluster is calculated independently. The dataset is then optimized by dynamic programming to find a sequence of subsets with low overall entropy according to imposed size limitations. The experimental evaluation shows that the proposed approach improves over current best alternatives in terms of accuracy, precision, recall, and F1-score metrics. Moreover, the proposed approach provides excellent interclass discrimination even for a large number of classes.

Face masks pose challenges for face recognition due to missing features. Current methods primarily rely on inpainting and reconstruction to improve recognition, but reconstructed images often lose identity and have low face similarity. This happens because the newly reconstructed features either come from other persons or are newly generated features. This paper proposes a method to improve the SSIM (Structural Similarity Index Measure) value, face recognition accuracy, and identity preservation by augmenting only the lower part of masked face images, rather than generating the entire face. It first analyzes several masked face images to detect the occluded area and learns the line between the visible and non-visible parts. Then it creates two datasets containing the upper part and lower parts of faces. A pre-trained CNN matches feature maps of the upper part of a query image with upper part dataset images to find a candidate image. Various techniques such as SURF are utilized to detect the geometrical property difference which is applied to the lower part of the candidate image to form the final full-face image. Our system’s performance was evaluated using LFW, CASIA-WebFace, AR, and FACES datasets. Accuracy, precision, recall, and F-1 score were calculated and compared with conventional methods. The proposed method enhanced recognition accuracy by 4-6% and significantly increased the SSIM value. It also demonstrated greater convenience, reduced runtime, and lower computational costs compared to existing methods.

In the contemporary digital marketplace, the proliferation of online consumer reviews has a pivotal influence on purchasing decisions. Concurrently, the prevalence of spurious reviews poses a substantial risk to the integrity of e-commerce, misleading consumers, and detrimentally impacting businesses. This paper delineates a pioneering methodology for the identification of counterfeit reviews, which is based on the combination of deep learning attributes and aspect-based analytical features. The main contribution of this research is (1) proposing an aspect fusion network based on the hierarchical attention mechanism to address the problem of multiple aspects of representing review content. The aspect fusion network can help select important aspect words and fuse aspect dictionaries with word-level attention weights. (2) We build a cardinality fusion model so that the heuristic can mitigate the negative impact of random weights and intervals on the auxiliary model. The methodology integrates advanced deep learning paradigms with aspect-based sentiment analysis to detect fraudulent reviews. Specifically, the approach encompasses a dual-method strategy: initially utilizing a Convolutional Neural Network (CNN) for the extraction of profound characteristics from review texts, followed by employing aspect-based sentiment analysis tools, including Part-of-Speech (PoS) tagging and GloVe embedding, for the distillation of aspectual features. Subsequently, these split sets of features are synergized and applied in the training of various classifier layers. Eextensive experiments have been conducted on six public review datasets contrasting the previous work on authenticity and aspect analysis. The effectiveness and performance of the proposed authenticity fusion model have been verified by the detailed analyses. The proposed model outperforms the competitors with remarkable improvement on both review authenticity and aspect analysis. This innovative approach was rigorously evaluated using a dataset of Amazon reviews that encompassed both authentic and counterfeit reviews. The empirical results demonstrate that our proposed method attains a remarkable accuracy rate of 97.73%, substantially surpassing existing state-of-the-art methodologies. The study posits that the strategic fusion of deep learning attributes and aspect-based features significantly enhances the efficacy of counterfeit review detection systems, presenting a formidable tool in the arsenal against e-commerce fraud.

In clinical routine, the assessment of myocardial viability is based on visual analysis of late gadolinium enhancement (LGE) sequences. This procedure remains subjective and insufficient, particularly in cases of improbable viability, where scar transmurality spans 25% to 75% of the myocardial segment total thickness. To address these challenges, this paper introduces a novel framework based on deep convolutional neural network (CNN) for objective and quantitative assessment of myocardial viability. The proposed method was validated on 73 patients with myocardial infarction and 10 healthy subjects. The initial stage involves the automatic quantification of regional myocardial wall thickness (MWT) in multiple radial directions to offer a detailed regional assessment compared to traditional techniques. This method is based on automatic segmentation of the left ventricle contours using U-Net. Afterwards, we proposed a novel protocol to automate the classification of myocardial segments into viable and nonviable classes. Additionally, we introduced MWT as a new key parameter for studying peri-infarct areas. Comparative study of our method to related works proves its superiority with an average mean absolute error (MAE) of $1.21~\pm ~1.00$ . Accurate quantification of MWT allowed the detection of myocardial segments desynchronization and the delimitation of infarction transmurality with an accuracy of 98.13%, a specificity of 99.09% and a sensitivity of 97.52% with (p_value

In order to improve patient outcomes, brain tumors—which are notorious for their catastrophic effects and short life expectancy, particularly in higher grades—need to be diagnosed accurately and treated with care. Patient survival chances may be hampered by incorrect medical procedures brought on by a brain tumor misdiagnosis. CNNs and computer-aided tumor detection systems have demonstrated promise in revolutionizing brain tumor diagnostics through the application of ML techniques. One issue in the field of brain tumor detection and classification is the dearth of non-invasive indication support systems, which is compounded by data scarcity. Conventional neural networks may cause problems such as overfitting and gradient vanishing when they use uniform filters in different visual settings. Moreover, these methods incur time and computational complexity as they train the model from scratch and extract the pertinent characteristics. This paper presents an InceptionV4 neural network architecture-based Transfer Learning-based methodology to address the shortcomings in brain tumor classification methods. The goal is to deliver precise diagnostic assistance while minimizing calculation time and improving accuracy. The model makes use of a dataset that contains 7022 MRI images that were obtained from figshare, the SARTAJ dataset, and Br35H, among other sites. The suggested InceptionV4 architecture improves its ability to categorize brain tumors into three groups and normal brain images by utilizing transfer learning approaches. The suggested InceptionV4 model achieves an accuracy rate of 98.7% in brain tumor classification, indicating the model’s remarkable performance. This suggests a noteworthy progression in the precision of diagnosis and computational effectiveness to support practitioners making decisions.

Ship detection holds a pivotal position in intelligent maritime transportation systems, ensuring the stable development of maritime engineering. However, the complex marine environment and practical deployment requirements pose numerous challenges to ship detection tasks. To address these limitations, this paper proposes a robust and real-time ship object detection method based on enhanced CNN (RRSD). Firstly, an efficient feature extraction module (EFEM) is proposed, which introduces a cross-stage stacked enhanced multi-head self-attention to reduce the computational complexity of the model while enhancing the network’s feature extraction capability. Next, a feature fusion network (FFM) is introduced to enhance the feature representation capability of small objects by effectively aggregating features of different scales in the backbone. To focus on the attention regions in the image, designing a lightweight enhanced attention module (EAM) to improve the model’s inference speed and classification ability. Additionally, by utilizing data augmentation techniques without any additional cost, the overall performance of the model is intuitively and efficiently improved. The proposed method is extensively experimented on a self-built ship detection dataset (SDD2024) through qualitative and quantitative experiments. The experimental results indicate that RRSD achieves a detection performance of 88.01mAP50 (mean average precision at an intersection over unionof 50) and 7.8ms, significantly outperforming object detection models of the same scale, and achieving a good trade-off between detection accuracy and speed. Furthermore, robustness experiments and embedded deployment tests demonstrate its good detection accuracy, fast speed, and satisfactory robustness in ship detection tasks, which basically meet the practical requirements of deployment and implementation.