Application Of Deep Learning Algorithms Research Articles

Comparison of deep learning-based models for detection of diseased trees using an image compression algorithm

The object of the research is the application of deep learning algorithms using an improved mathematical lossless image compression method for recognizing and identifying dead trees in aerospace images. The main problem that has been solved is the archiving of images due to their large volume on disk and the possibility of their further processing by deep learning methods such as convolutional and capsule neural networks, which have shown high efficiency and accuracy in image recognition and classification tasks using the proposed new image compression method. The article presents a comparative analysis of the performance of three YOLO (You Only Look Once) models with different types of architectures, such as YOLOv5, YOLOv7 and YOLOv8, to assess the effectiveness of their work for the task of recognizing aerospace tree images obtained from satellites, drones, and aircrafts. Comprehensive analysis of YOLO models presents that model YOLO v8 turned out to be most effective with a positive accuracy of 88.2 %, a recall of 77.4 %, and a mAP50 score of 87.2 %. Moreover, the average detection time was only 0.052 seconds for each image, even though the model size remains very small – 21.5 MB. These results suggest a much better usage of time and precise identification of dead trees, and classified targets with high efficiency. From the research, there is significant prospects of global forest management especially on forest reduction and protection of ecosystems through accurate assessment on the health of forestry. The proposed approach is universal and can be used in real life conditions, providing a good compromise of the speed, accuracy and resources required for forest monitoring and management

Construction of English corpus oral instant translation model based on information security and deep learning

In order to improve the quality of translation, avoid translation ambiguity and accurately present the content of the source language, supported by the concept of deep learning and guaranteed by information security, an instant oral translation model is constructed for English corpus. The aim of this study is to enhance the efficiency and accuracy of oral translation systems through the application of deep learning algorithms. Specifically, we employ a sample training mechanism tailored to the unique characteristics of oral translation, allowing for separate training of system interaction and translation data. Furthermore, by redesigning the interaction hardware, this research comprehensively redefines the hardware structure of the translation system, marking a significant step towards improving the usability and performance of such systems. After obtaining and processing effective security sensitive information, language resources are managed by using database management system, which fundamentally improves the level of network information security. The performance of the existing oral automatic translation system (Test Group 1) and the system designed in this paper (Test Group 2) is tested by experiments, and the results are as follows: (1) The translation system designed here has better interactive performance, and it is better than Test Group 1. (2) The adaptive index value of Test Group 1 is 1, and that of Test Group 2 is 0.5, which proves that the adaptive ability of system algorithm of Test Group 2 is better than that of Test Group 1. (3) When comparing the translation speed, the translation time of Test Group 2 is only 70.7 s, while that of Test Group 1 is 130.6 s, so the proposed translation system is obviously superior to that of Test Group 1.

A multi-objective hierarchical deep reinforcement learning algorithm for connected and automated HEVs energy management

Connected and autonomous vehicles have offered unprecedented opportunities to improve fuel economy and reduce emissions of hybrid electric vehicle (HEV) in vehicular platoons. In this context, a hierarchical control strategy is put forward for connected HEVs. Firstly, we consider a deep deterministic policy gradient (DDPG) algorithm to compute the optimized vehicle speed using a trained optimal policy via vehicle-to-vehicle communication in the upper level. A multi-objective reward function is introduced, integrating vehicle fuel consumption, battery state-of-the-charge, emissions, and vehicle car-following objectives. Secondly, an adaptive equivalent consumption minimization strategy is devised to implement vehicle-level torque allocation in the platoon. Two drive cycles, HWFET and human-in-the-loop simulator driving cycles are utilized for realistic testing of the considered platoon energy management. It is shown that DDPG runs the engine more efficiently than the widely-implemented Q-learning and deep Q-network, thus showing enhanced fuel savings. Further, the contribution of this paper is to speed up the higher-level vehicular control application of deep learning algorithms in the connected and automated HEV platoon energy management applications.

A New Method for Scoliosis Screening Incorporating Deep Learning With Back Images.

Retrospective observational study. Scoliosis is commonly observed in adolescents, with a world0wide prevalence of 0.5%. It is prone to be overlooked by parents during its early stages, as it often lacks overt characteristics. As a result, many individuals are not aware that they may have scoliosis until the symptoms become quite severe, significantly affecting the physical and mental well-being of patients. Traditional screening methods for scoliosis demand significant physician effort and require unnecessary radiography exposure; thus, implementing large-scale screening is challenging. The application of deep learning algorithms has the potential to reduce unnecessary radiation risks as well as the costs of scoliosis screening. The data of 247 scoliosis patients observed between 2008 and 2021 were used for training. The dataset included frontal, lateral, and back upright images as well as X-ray images obtained during the same period. We proposed and validated deep learning algorithms for automated scoliosis screening using upright back images. The overall process involved the localization of the back region of interest (ROI), spinal region segmentation, and Cobb angle measurements. The results indicated that the accuracy of the Cobb angle measurement was superior to that of the traditional human visual recognition method, providing a concise and convenient scoliosis screening capability without causing any harm to the human body. The method was automated, accurate, concise, and convenient. It is potentially applicable to a wide range of screening methods for the detection of early scoliosis.

Application of Deep Learning Algorithms Based on the Multilayer YOLOv8 Neural Network to Identify Fungal Keratitis

Application of Deep Learning Algorithms Based on the Multilayer YOLOv8 Neural Network to Identify Fungal Keratitis

An In-Depth Review of Predictive Methods for Oil and Gas Applications

The oil and gas industry aims to optimize production, reduce costs, and increase efficiency. Predictive models have gained popularity as potential solutions in recent years. In predictive models, machine learning algorithms analyse oil and gas operations data and forecast future performance. This paper examines the current state of predictive modelling in the oil and gas industry with the objective to systematically review and analyse current research on the predictive modelling in the oil and gas industry. The paper begins by highlighting the sub-fields and datasets in the oil and gas industry that used recent machine learning methods for predictive modelling. Additionally, literature from the Scopus and Web of Science indexes was reviewed. This study assessed recent approaches for oil and gas industry in predictive applications for papers published up until December 2022. The findings identify several advantages and disadvantages that can be used as guidelines to effectively implement predictive modelling in the oil and gas industry. It includes challenges on the requirement for accurate and reliable data, the development of appropriate algorithms, and the integration of predictive models into existing workflows. In addition, the finding highlights the growing application of deep learning algorithms for various tasks as one of the major trends. From the analysis of the state-of-the-art in predictive techniques, it is necessary first to survey the landscape of existing predictive analytics approaches and their methods to employ in oil and gas prediction.

Developing deep learning surrogate models for digital twins in mineral processing – A case study on data-driven multivariate multistep forecasting

The escalating demand for environmental and social sustainability underscores the critical need for large industries such as mining and metallurgy to function optimally. Achieving optimal operation of a mineral processing plant requires timely access to information and rapid implementation of corrective actions. Essential information, such as current and future production, plays a pivotal role in enhancing process efficiency by facilitating adjustments in key aspects, including feed, control and operating variables. This study delves into the application of deep learning algorithms to forecast the gold concentrate grade of a flotation process half an hour in advance, a valuable tool for short-term decision-making, e.g., with regard to optimising the operation and process control. Multivariate time series data from a real gold processing plant was acquired through the process instrumentation and control system and augmented with simulated data at 5-minute intervals over 236 days. Surrogate models based on LSTM, GRU and CNN architectures were developed and analysed for predicting the gold grade one to six steps ahead. Evaluation of the testing dataset indicates that the CNN exhibits superior performance, achieving an average error reduction of 36 %, 45 % and 35 % for the performance metrics RMSE, MAE and MAPE respectively compared to the baseline model. These findings reveal the high potential of deep learning surrogate methods for accurately forecasting process outputs, enabling effective data-driven process control and optimisation strategies.

Human gender estimation from CT images of skull using deep feature selection and feature fusion

This research endeavors to prognosticate gender by harnessing the potential of skull computed tomography (CT) images, given the seminal role of gender identification in the realm of identification. The study encompasses a corpus of CT images of cranial structures derived from 218 male and 203 female subjects, constituting a total cohort of 421 individuals within the age bracket of 25 to 65 years. Employing deep learning, a prominent subset of machine learning algorithms, the study deploys convolutional neural network (CNN) models to excavate profound attributes inherent in the skull CT images. In pursuit of the research objective, the focal methodology involves the exclusive application of deep learning algorithms to image datasets, culminating in an accuracy rate of 96.4%. The gender estimation process exhibits a precision of 96.1% for male individuals and 96.8% for female individuals. The precision performance varies across different selections of feature numbers, namely 100, 300, and 500, alongside 1000 features without feature selection. The respective precision rates for these selections are recorded as 95.0%, 95.5%, 96.2%, and 96.4%. It is notable that gender estimation via visual radiography mitigates the discrepancy in measurements between experts, concurrently yielding an expedited estimation rate. Predicated on the empirical findings of this investigation, it is inferred that the efficacy of the CNN model, the configurational intricacies of the classifier, and the judicious selection of features collectively constitute pivotal determinants in shaping the performance attributes of the proposed methodology.

Application of High-Spatial-Resolution Imagery and Deep Learning Algorithms to Spatial Allocation of Urban Parks’ Supply and Demand in Beijing, China

The development of green spaces in urban parks can significantly enhance the quality of the urban and ecological environment. This paper utilizes 2021 Gaofen-7 (GF-7) satellite remote sensing images as its primary data source and uses deep learning algorithms for the precise extraction of the green space coverage within Beijing’s fifth ring road. It also incorporates the park points of interest (POI) information, road data, and other auxiliary data to extract green park space details. The analysis focuses on examining the relationship between supply and demand in the spatial allocation of green park spaces from an accessibility perspective. The main findings are as follows: (1) The application of deep learning algorithms improves the accuracy of green space extraction by 10.68% compared to conventional machine methods. (2) The distribution of parks and green spaces within the fifth ring road of Beijing is uneven, showing a clear pattern of “more in the north and less in the south”. The accessibility within a five-minute service radius achieves a coverage rate of 46.65%, with a discernible blind zone in the southeast. (3) There is an imbalance in the per capita green space location entropy within the fifth ring road of Beijing, there is a big difference in per capita green space location entropy (44.19), and social fairness needs to be improved. The study’s outcomes unveil the intricate relationship between service capacity and spatial allocation, shedding light on the supply and demand dynamics of parks and green spaces within Beijing’s fifth ring road. This insight will contribute to the construction of ecologically sustainable and aesthetically pleasing living spaces in modern megacities.

Application of deep learning algorithms in classification and localization of implant cutout for the postoperative hip.

This study aims to explore the feasibility of employing convolutional neural networks for detecting and localizing implant cutouts on anteroposterior pelvic radiographs. The research involves the development of two Deep Learning models. Initially, a model was created for image-level classification of implant cutouts using 40191 pelvic radiographs obtained from a single institution. The radiographs were partitioned into training, validation, and hold-out test datasets in a 6/2/2 ratio. Performance metrics including the area under the receiver operator characteristics curve (AUROC), sensitivity, and specificity were calculated using the test dataset. Additionally, a second object detection model was trained to localize implant cutouts within the same dataset. Bounding box visualizations were generated on images predicted as cutout-positive by the classification model in the test dataset, serving as an adjunct for assessing algorithm validity. The classification model had an accuracy of 99.7%, sensitivity of 84.6%, specificity of 99.8%, AUROC of 0.998 (95% CI: 0.996, 0.999) and AUPRC of 0.774 (95% CI: 0.646, 0.880). From the pelvic radiographs predicted as cutout-positive, the object detection model could achieve 95.5% localization accuracy on true positive images, but falsely generated 14 results from the 15 false-positive predictions. The classification model showed fair accuracy for detection of implant cutouts, while the object detection model effectively localized cutout. This serves as proof of concept of using a deep learning-based approach for classification and localization of implant cutouts from pelvic radiographs.

Prediction of Mismatch Repair Status in Endometrial Cancer from Histological Slide Images Using Various Deep Learning-Based Algorithms.

The application of deep learning algorithms to predict the molecular profiles of various cancers from digital images of hematoxylin and eosin (H&E)-stained slides has been reported in recent years, mainly for gastric and colon cancers. In this study, we investigated the potential use of H&E-stained endometrial cancer slide images to predict the associated mismatch repair (MMR) status. H&E-stained slide images were collected from 127 cases of the primary lesion of endometrial cancer. After digitization using a Nanozoomer virtual slide scanner (Hamamatsu Photonics), we segmented the scanned images into 5397 tiles of 512 × 512 pixels. The MMR proteins (PMS2, MSH6) were immunohistochemically stained, classified into MMR proficient/deficient, and annotated for each case and tile. We trained several neural networks, including convolutional and attention-based networks, using tiles annotated with the MMR status. Among the tested networks, ResNet50 exhibited the highest area under the receiver operating characteristic curve (AUROC) of 0.91 for predicting the MMR status. The constructed prediction algorithm may be applicable to other molecular profiles and useful for pre-screening before implementing other, more costly genetic profiling tests.

Satellite remote sensing image classification based on multiple deep learning algorithms

This paper explores the use of satellite remote sensing technology to observe and measure the earth's surface and classify remote sensing images using various deep learning algorithms. By importing AlexNet, VggNet, GoogleNet and MobileNet models, 90% of the data are randomly selected for training and 10% for testing, and the changes in the loss and accuracy of the validation set during training are recorded, as well as the accuracy of the best round of epochs for training.During the training process, the validation set's loss gradually decreased and converged, and the accuracy of the validation set gradually increased and stabilised. The results show that all four models are able to classify remote sensing images well, among which the highest accuracy is MobileNet model, which reaches 99.8%, and Googlenet model is the second one, which reaches 97.9%.AlexNet and VggNet models also have higher accuracy, which are 93.3% and 95.4%, respectively. Overall, the results of this paper provide an effective solution for satellite remote sensing image classification and provide a reference for the application of deep learning algorithms in remote sensing.

Research on Methods of Micro-target Detection in Various Application Scenarios

This paper provides a comprehensive introduction and comparison of the research on small target detection algorithms in different application scenarios and systematically explores the application of deep learning algorithms for small target detection in remote sensing imagery, visible imagery, and infrared imagery. In remote sensing images, the Deconvolution R-CNN algorithm is used to improve the accuracy of small target detection by adding a deconvolution layer. In the field of visible images, a Gaussian Mixture Model (GMM)-based approach is introduced to achieve fast detection of small targets with SIFT features and a GMM target detector. In infrared images, a comprehensive algorithm adapted to low signal-to-noise ratio and complex background is proposed to automatically detect tiny targets through background suppression and threshold selection optimization. In this paper, we compare and analyze the performance of these methods in different scenarios, and propose directions for future research, including expanding the training dataset, applying super-resolution methods to improve the image resolution, and enhancing the universality of the algorithms. This paper provides a comprehensive overview of research in the field of small target detection, as well as an outlook on future research directions, which provides a useful reference for the further development of the field.

Enhancing green sea turtle (Chelonia mydas) conservation for tourists at Little Liuqiu island, Taiwan: Application of deep learning algorithms

Observing marine life has emerged as a pivotal catalyst for the growth of the Blue Economy. Yet, overzealous and recurrent observations may exert undue stress on marine creatures, thereby complicating the sustainable management of marine assets. After the easing of Covid-19 pandemic restrictions, a notable influx of tourists to Little Liuqiu Island, Taiwan puts considerable stress on its green sea turtle population. This surge intensified incidents of illegal harassment, triggering concerns from both the local community and administrative bodies. The prevailing challenge is to ensure tourists observe these turtles with respect, refraining from behaviors such as touching or pursuing them. Addressing this, our study harnesses deep learning algorithms to equip ecotourism operators and tourists with tools to detect green sea turtles across diverse coastal terrains, reinforcing conservation efforts. Our analysis scrutinized object detection AI models, namely YOLOv3, YOLOv5s, and YOLOv5l. Fieldwork was undertaken on the island to gather ample training images, capturing elusive green sea turtles in various settings, from coastline strolls to drone imagery. Supplemental images sourced from local social media platforms were later added. Contrary to expectations, we found that merely expanding the training dataset did not guarantee improved outcomes. Instead, the variance in image content, considering distances, angles, and turtle appearances, played a pivotal role in enhancing model precision. Through our experimentation, the streamlined YOLOv5s model consistently eclipsed its more complex counterparts in performance. An AI service, underpinned by the YOLOv5s model, has been launched to distinguish between green sea turtle types for tourist-focused conservation initiatives. Future iterations will incorporate user feedback to refine accuracy. Our research breaks new ground, spotlighting the intricacies of gathering natural environment data, pinpointing optimal AI models, and evaluating their practical implications for green sea turtle conservation.

Breast Cancer Classification Utilizing Deep Learning Techniques on Medical Images: A Comprehensive Review

Breast cancer is a major public health concern, affecting millions of women worldwide. Good patient outcomes and a successful course of therapy depend on a prompt and accurate diagnosis. The application of deep learning algorithms to the breast cancer classification problem has yielded some encouraging results; this might open the way for more rapid and accurate diagnostics. This article examines the many deep learning (DL) approaches taken to date for BC classification tasks, outlining each one's strengths, weaknesses, and current challenges. This study discusses several DL algorithms like convolutional neural networks, multi-layer neural networks, and autoencoders that analyze histopathological pictures, mammograms, and other forms of medical imaging. Furthermore, we investigate the interpretability & explainability features of DL models as they pertain to BC diagnoses, drawing attention to the need for reliable decision-making tools for medical practitioners. Throughout the review, we identify challenges and potential biases in current research. Finally, we provide an outlook on the future directions of DL in BC classification, focusing on promising research areas. By highlighting the achievements and gaps in the existing literature, this review aims to inspire further advancements in DL-based BC diagnosis, eventually resulting in better healthcare and reliable diagnosis.

Application of deep learning algorithms in predicting the exchange rate of Chinese yuan against the US dollar

This paper mainly studies the feasibility of using deep learning algorithms to predict the exchange rate between the Chinese yuan and the US dollar. Firstly, this study chooses the Long Short Term Memory Network (LSTM) as the main algorithm, which is a deep learning model suitable for processing time series data. Then, the study collects a large amount of data, including historical exchange rate data, macroeconomic indicators, and political events that may affect exchange rates. These data provide sufficient information for the model to learn and predict exchange rate fluctuations. The study found that the deep learning model can accurately predict the volatility of the RMB US dollar exchange rate by contrasting the calculated results of the pattern with effective exchange rate record, and its predicted results have high stability. Deep learning has great potential in the field of exchange rate forecasting, providing more accurate predictive information for financial transactions.

Sample Augmentation Method for Side-Scan Sonar Underwater Target Images Based on CBL-sinGAN

The scarcity and difficulty in acquiring Side-scan sonar target images limit the application of deep learning algorithms in Side-scan sonar target detection. At present, there are few amplification methods for Side-scan sonar images, and the amplification image quality is not ideal, which is not suitable for the characteristics of Side-scan sonar images. Addressing the current shortage of sample augmentation methods for Side-scan sonar, this paper proposes a method for augmenting single underwater target images using the CBL-sinGAN network. Firstly, considering the low resolution and monochromatic nature of Side-scan sonar images while balancing training efficiency and image diversity, a sinGAN network is introduced and designed as an eight-layer pyramid structure. Secondly, the Convolutional Block Attention Module (CBAM) is integrated into the network generator to enhance target learning in images while reducing information diffusion. Finally, an L1 loss function is introduced in the network discriminator to ensure training stability and improve the realism of generated images. Experimental results show that the accuracy of shipwreck target detection increased by 4.9% after training with the Side-scan sonar sample dataset augmented by the proposed network. This method effectively retains the style of the images while achieving diversity augmentation of small-sample underwater target images, providing a new approach to improving the construction of underwater target detection models.

Classification of hazelnut varieties based on bigtransfer deep learning model

Hazelnut is an agricultural product that contributes greatly to the economy of the countries where it is grown. The human factor plays a major role in hazelnut classification. The typical approach involves manual inspection of each sample by experts, a process that is both labor-intensive and time-consuming, and often suffers from limited sensitivity. The deep learning techniques are extremely important in the classification and detection of agricultural products. Deep learning has great potential in the agricultural sector. This technology can improve product quality, increase productivity, and offer farmers the ability to classify and detect their produce more effectively. This is important for sustainability and efficiency in the agricultural industry. In this paper aims to the application of deep learning algorithms to streamline hazelnut classification, reducing the need for manual labor, time, and cost in the sorting process. The study utilized hazelnut images from three different varieties: Giresun, Ordu, and Van, comprising a dataset of 1165 images for Giresun, 1324 for Ordu, and 1138 for Van hazelnuts. This dataset is an open-access dataset. In the study, experiments were carried out on the determination of hazelnut varieties with BigTransfer (BiT)-M R50 × 1, BiT-M R101 × 3 and BiT-M R152 × 4 models. Deep learning models, including big transfer was employed for classification. The classification task involved 3627 nut images and resulted in a remarkable accuracy of 99.49% with the BiT-M R152 × 4 model. These innovative methods can also lead to patentable products and devices in various industries, thereby boosting the economic value of the country.

Research on the application of deep learning algorithms to PCB defect detection

In today's electronics industry, Printed Circuit Boards play a crucial role in providing the layout for circuit components and conductive traces in nearly all electronic devices. The quality of components soldered onto Printed Circuit Boards directly impacts product performance. To ensure the performance of electronic devices, Printed Circuit Boards defect detection based on deep learning algorithms has become a pivotal technology in the defect inspection process within the electronics industry. However, the application of deep learning algorithms in this context faces several challenges. These challenges include difficulties in acquiring Printed Circuit Boards defect datasets, limited generalization capability in Printed Circuit Boards defect detection, and slow and low-quality Printed Circuit Boards image stitching processes. To enhance researchers' understanding of deep learning-based Printed Circuit Boards defect detection, this paper analyzes the challenges associated with deep learning in the Printed Circuit Boards defect detection process and proposes several viable solutions. In conclusion, this paper provides insights into the future of deep learning-based Printed Circuit Boards defect detection.

Multi-modal physiological time-frequency feature extraction network for accurate sleep stage classification

Sleep stage classification is essential for clinical disease diagnosis and sleep quality assessment. Most of the existing methods for sleep stage classification are based on single-channel or single-modal signal, and extract features using a single-branch, deep convolutional network, which not only hinders the capture of the diversity features related to sleep and increase the computational cost, but also has a certain impact on the accuracy of sleep stage classification. To solve this problem, this paper proposes an end-to-end multi-modal physiological time-frequency feature extraction network (MTFF-Net) for accurate sleep stage classification. First, multi-modal physiological signal containing electroencephalogram (EEG), electrocardiogram (ECG), electrooculogram (EOG) and electromyogram (EMG) are converted into two-dimensional time-frequency images containing time-frequency features by using short time Fourier transform (STFT). Then, the time-frequency feature extraction network combining multi-scale EEG compact convolution network (Ms-EEGNet) and bidirectional gated recurrent units (Bi-GRU) network is used to obtain multi-scale spectral features related to sleep feature waveforms and time series features related to sleep stage transition. According to the American Academy of Sleep Medicine (AASM) EEG sleep stage classification criterion, the model achieved 84.3% accuracy in the five-classification task on the third subgroup of the Institute of Systems and Robotics of the University of Coimbra Sleep Dataset (ISRUC-S3), with 83.1% macro F1 score value and 79.8% Cohen's Kappa coefficient. The experimental results show that the proposed model achieves higher classification accuracy and promotes the application of deep learning algorithms in assisting clinical decision-making.