Leveraging Deep Learning Research Articles

DDC: Deep Distribution Classifier, A Convolutional Neural Network-based Approach for Identifying Data Distributions

In domains such as the stock market and manufacturing, there’s a growing demand for faster and more accurate data distribution identification methods due to the rapid generation of vast volumes of data, highlighting the need for enhanced real-time decision-making capabilities. Traditional methods of identifying data distributions often rely on manual inspection, limited statistical tests and time-consuming analysis, leading to inefficiencies and inaccuracies in classification. In this scenario, the presented research offers a novel approach leveraging Deep Learning (DL) models to automate the process. The presented methodology also enables faster and more accurate identification of data distributions by the generation of synthetic data points and training of the DL model for identifying different distribution types. The primary objective of this study is to develop a DL model that categorizes data points into specific distributions based on an input dataset. Moreover, for model training and evaluation, a total of 1000 datasets are generated,each comprising 1000 data points. The study considers five distributions (Normal, Uniform, Exponential, Log-normal and Beta distribution), with 200 datasets generated (with randomly selected parameters) for each distribution. In the study, the DL model is trained first, and later, the model is evaluated on a separate test (unseen) dataset. Then, its performance in classifying the distributions is assessed based on metrics such as accuracy and loss. The study results demonstrate the effectiveness of the proposed approach in accurately classifying the distribution of data points, providing valuable insights into the application of DL for distribution classification tasks. The proposed method enhances scalability, robustness and efficiency by harnessing the power of convolutional neural networks and advanced preprocessing techniques.

Reimagining Accessibility: Leveraging Deep Learning in Smartphone Applications to Assist Visually Impaired People Indoor Object Distance Estimation

Every other aspect of life is organized around the sight. A person with vision impairment suffers severely from independent mobility and quality of life. The proposed framework combines mobile deep learning with distance estimation algorithms to detect and classify indoor objects with estimated distance in real-time during indoor navigation. The user, wearing the device with a lanyard or holding it in a way that positions the camera forward, identifies in real-time surroundings indoor objects with estimated distance and voice commentary. Moreover, the mobile framework provides an estimated distance to the obstacles and suggests a safe navigational path through voice-guided feedback. By harnessing the power of deep learning in a mobile setting, this framework aims to improve the independence of visually impaired individuals by facilitating them a higher degree of independence in indoor navigation. This study's proposed mobile object detection and distance estimation framework achieved 99.75% accuracy. This research contributes by leveraging mobile deep learning with identifying objects in real-time, classification and distance estimation, a state-of-the-art approach to use the latest technologies to enhance indoor mobility challenges faced by visually impaired people.

An Automated Quantitative Methodology for Computing Gravel Parameters in Imaging Logging Leveraging Deep Learning: A Case Analysis of the Baikouquan Formation within the Mahu Sag

Gravels are widely distributed in the Baikouquan formation in the Mabei area of the Junggar Basin. However, conventional logging methods cannot quantitatively characterize gravel development, which limits the identification of lithology, structure, and sedimentary facies in this region. This study proposes a new method for automatically identifying gravels from electric imaging images and calculating gravel parameters utilizing the salient object detection (SOD) network. Firstly, a SOD network model (U2-Net) was constructed and trained using electric imaging data from the Baikouquan formation at the Mahu Sag. The blank strips in the images were filled using the U-Net convolutional neural network model. Sample sets were then prepared, and the gravel areas were labeled in the electric imaging images with the Labelme software in combination with image segmentation and human–machine interaction. These sample sets were used to train the network model, enabling the automatic recognition of gravel areas and the segmentation of adhesive gravel regions in the electric imaging images. Based on the segmented gravel results, quantitative evaluation parameters such as particle size and gravel quantity were accurately calculated. The method’s validity was confirmed through validation sets and actual data. This approach enhances adhesive area segmentation’s accuracy and processing speed while effectively reducing human error. The trained network model demonstrated an average absolute error of 0.0048 on test sets with a recognition accuracy of 83.7%. This method provides algorithmic support for the refined evaluation of glutenite reservoir logging.

Bird Sound Classification : Leveraging Deep Learning for Species Identification

Birds are meaningful to a wide audience including the public. They live in almost every type of environment and in almost every niche (place or role) within those environments. The monitoring of species diversity and migration is important for almost all conservation efforts. The analysis of long-term audio data is vital to support those efforts but relies on complex algorithms that need to adapt to changing environmental conditions. Convolutional neural networks (CNNs) are powerful toolkits of machine learning that have proven efficient in the field of image processing and sound recognition. In this paper, a CNN system classifying bird sounds is presented and tested through different configurations and hyperparameters. The MobileNet pre-trained CNN model is finetuned using a dataset acquired from the Xeno-canto bird song sharing portal, which provides a large collection of labeled and categorized recordings. Spectrograms generated from the downloaded data represent the input of the neural network. The attached experiments compare various configurations including the number of classes (bird species) and the color scheme of the spectrograms. Results suggest that choosing a color map in line with the images the network has been pre-trained with provides a measurable advantage. The presented system is viable only for a low number of classes.

Combating Misinformation: Leveraging Deep Learning for Hoax Detection in Indonesian Political Social Media

The rampant spread of hoax news in social media, especially in the political domain, poses a significant challenge that requires immediate attention. To address this issue, automatic hoax news detection using machine learning-based artificial intelligence has emerged as a promising approach. With the approaching presidential election in Indonesia in 2024, the need for effective detection methods becomes even more pressing.This research focuses on proposing an efficient deep learning model for detecting political hoax news on Indonesian social media. Word2vec feature representation and three deep learning models – LSTM, CNN, and Hybrid CNN-LSTM – are evaluated to determine the most effective approach. Experimental results reveal that the CNN-LSTM hybrid model outperforms the others, achieving an accuracy of 96% in detecting hoax news on Indonesian social media in the political domain. By leveraging state-of-the-art deep learning techniques, particularly the CNN-LSTM hybrid model, this study contributes to the advancement of hoax news detection in Indonesia's political landscape. The findings underscore the importance of utilizing sophisticated machine learning methods to combat the spread of misinformation, particularly during crucial political events such as elections.

Road Network Mapping from Multispectral Satellite Imagery: Leveraging Deep Learning and Spectral Bands

Abstract. Updating road networks in rapidly changing urban landscapes is an important but difficult task, often challenged by the complexity and errors of manual mapping processes. Traditional methods that primarily use RGB satellite imagery struggle with obstacles in the environment and varying road structures, leading to limitations in global data processing. This paper presents an innovative approach that utilizes deep learning and multispectral satellite imagery to improve road network extraction and mapping. By exploring U-Net models with DenseNet backbones and integrating different spectral bands we apply semantic segmentation and extensive post-processing techniques to create georeferenced road networks. We trained two identical models to evaluate the impact of using images created from specially selected multispectral bands rather than conventional RGB images. Our experiments demonstrate the positive impact of using multispectral bands, by improving the results of the metrics Intersection over Union (IoU) by 6.5%, F1 by 5.4%, and the newly proposed relative graph edit distance (relGED) and topology metrics by 2.2% and 2.6% respectively.

Car Accident Detection and Alert System Leveraging Deep Learning

This paper presents a novel approach to address the critical issue of car accidents through an advanced Accident Detection and Alerting System (ADAS) empowered by cutting-edge technologies in deep learning, computer vision, and communication APIs. With road safety being a paramount concern worldwide, timely detection of car accidents followed by swift alerting mechanisms can significantly reduce the severity of accidents and save lives. The proposed system architecture incorporates the Ultralytics framework, Convolutional Neural Networks (CNN), You Only Look Once version 8 (YOLOv8), OpenCV, and Twilio API. Initially, YOLOv8 is employed to perform real-time detection of vehicles and potential accident scenarios from live video feeds obtained from roadside cameras or vehicle-mounted cameras. Subsequently, CNN models are utilized for comprehensive analysis of detected objects to discern potential accidents based on predetermined criteria, including sudden deceleration, collision patterns, and anomalous vehicle movements. Upon successful detection of an accident, the system triggers an alert mechanism via the Twilio API, which delivers instantaneous notifications to emergency services, nearby medical facilities, and designated contacts of the involved parties. These alerts are enriched with crucial details such as the accident location, vehicle types involved, and severity assessments derived from CNN analyses. Furthermore, the integration of OpenCV facilitates preprocessing tasks such as noise reduction, image enhancement, and object tracking, thereby enhancing the system's robustness and accuracy in diverse environmental conditions and lighting scenarios. The proposed ADAS offers numerous benefits, including real-time accident detection, automated alerting of emergency services, and minimized response times, ultimately contributing to mitigating the severity of accidents and preserving lives. By harnessing the power of deep learning algorithms, advanced computer vision techniques, and seamless communication APIs, this system represents a significant step towards fostering safer road environments within smart cities and transportation networks.

Differential Diagnosis of OKC and SBC on Panoramic Radiographs: Leveraging Deep Learning Algorithms.

This study aims to determine whether it can distinguish odontogenic keratocyst (OKC) and simple bone cyst (SBC) based solely on preoperative panoramic radiographs through a deep learning algorithm. (1) Methods: We conducted a retrospective analysis of patient data from January 2018 to December 2022 at Pusan National University Dental Hospital. This study included 63 cases of OKC confirmed by histological examination after surgical excision and 125 cases of SBC that underwent surgical curettage. All panoramic radiographs were obtained utilizing the Proline XC system (Planmeca Co., Helsinki, Finland), which already had diagnostic data on them. The panoramic images were cut into 299 × 299 cropped sizes and divided into 80% training and 20% validation data sets for 5-fold cross-validation. Inception-ResNet-V2 system was adopted to train for OKC and SBC discrimination. (2) Results: The classification network for diagnostic performance evaluation achieved 0.829 accuracy, 0.800 precision, 0.615 recall, and a 0.695 F1 score. (4) Conclusions: The deep learning algorithm demonstrated notable accuracy in distinguishing OKC from SBC, facilitated by CAM visualization. This progress is expected to become an essential resource for clinicians, improving diagnostic and treatment outcomes.

Oil Palm Bunch Ripeness Classification and Plantation Verification Platform: Leveraging Deep Learning and Geospatial Analysis and Visualization

Oil palm cultivation thrives as a prominent agricultural endeavor within the southern region of Thailand, where the country ranks third globally in production, following Malaysia and Indonesia. The assessment of oil palm bunch ripeness serves various purposes, notably in determining purchasing prices, pre-harvest evaluations, and evaluating the impacts of disasters or low market prices. Presently, two predominant methods are employed for this assessment, namely human evaluation, and machine learning for ripeness classification. Human assessment, while boasting high accuracy, necessitates the involvement of farmers or experts, resulting in prolonged processing times, especially when dealing with extensive datasets or dispersed fields. Conversely, machine learning, although capable of accurately classifying harvested oil palm bunches, faces limitations concerning its inability to process images of oil palm bunches on trees and the absence of a platform for on-tree ripeness classification. Considering these challenges, this study introduces the development of a classification platform leveraging machine learning (deep learning) in conjunction with geospatial analysis and visualization to ascertain the ripeness of oil palm bunches while they are still on the tree. The research outcomes demonstrate that oil palm bunch ripeness can be accurately and efficiently classified using a mobile device, achieving an impressive accuracy rate of 99.89% with a training dataset comprising 8779 images and a validation accuracy of 96.12% with 1160 images. Furthermore, the proposed platform facilitates the management and processing of spatial data by comparing coordinates derived from images with oil palm plantation data obtained through crowdsourcing and the analysis of cloud or satellite images of oil palm plantations. This comprehensive platform not only provides a robust model for ripeness assessment but also offers potential applications in government management contexts, particularly in scenarios necessitating real-time information on harvesting status and oil palm plantation conditions.

IoT Threat Mitigation: Leveraging Deep Learning for Intrusion Detection

The growth of smart gadgets connected via the Internet of Things (IoT) in today’s modern technology landscape has substantially improved our everyday lives. However, this convenience is juxtaposed with a concomitant surge in cyber threats capable of compromising the integrity of these interconnected systems. Conventional intrusion detection systems (IDS) prove inadequate for IoT due to the unique challenges they present. We propose and evaluate an intrusion detection system (IDS) based on a hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model in this paper. The model is designed to capture both temporal and spatial patterns in network data, offering a robust solution for detecting malicious activities within IoT environments. The CNN-LSTM model displayed excellent accuracy, reaching 98% in both multi-class and binary classifications when trained on the UNSW-NB15 dataset. Furthermore, we explore the real-world applicability of the model through testing on Raspberry Pi, showcasing its effectiveness in IoT scenarios. The system is augmented with alert mechanisms, promptly notifying relevant parties upon intrusion detection. Our findings highlight the CNN-LSTM model's efficacy in strengthening IoT network security.

Deep-learning framework for optimal selection of soil sampling sites

Soil sampling is one of the most fundamental agricultural processes. The selection of soil sampling locations within a field plays a crucial role in soil-health analysis. The current soil-sampling techniques produce random samples across a field, which are not representative of the field conditions. This work leverages the recent advancements of deep learning in image processing for finding optimal locations that present the important characteristics of a field. The data for training are collected at different fields in local farms with five features: aspect, flow accumulation, slope, NDVI (normalized difference vegetation index), and yield. The soil sampling dataset is challenging because of the unbalance between the number of pixels that represent soil-sampling spots and the number of pixels that represent the background area. In this work, we approach the problem with two methods, the first approach involves utilizing a state-of-the-art model with the convolutional neural network (CNN) backbone, while the second is to innovate a deep-learning design grounded in the concepts of transformer and self-attention. Our framework is constructed with an encoder–decoder architecture with the self-attention mechanism as the backbone. In the encoder, the self-attention mechanism is the key feature extractor, which produces feature maps. In the decoder, we introduce atrous convolution networks to concatenate, fuse the extracted features, and then export the optimal locations for soil sampling. Currently, the model has achieved impressive results on the testing dataset, with a mean accuracy of 99.52%, a mean Intersection over Union (IoU) of 57.35%, and a mean Dice Coefficient of 71.47%, while the performance metrics of the state-of-the-art CNN-based model are 66.08%, 3.85%, and 1.98%, respectively. This indicates that our proposed model outperforms the CNN-based method on the soil-sampling dataset. To the best of our knowledge, our work is the first to provide a soil-sampling dataset with multiple attributes and leverage deep learning techniques to enable the automatic selection of soil-sampling sites. This work lays a foundation for novel applications of data science and machine-learning technologies to solve other emerging agricultural problems.

Predicting Cyclone Michaung's Wrath: Leveraging Deep Learning Technique for Intensity and Track Forecasting

Tropical cyclones pose significant threats to coastal regions, necessitating accurate and timely monitoring systems for effective disaster management. In this study, Convolutional Neural Networks (CNN) is used to identify and track the tropical cyclone Michaung using a comprehensive dataset comprising INSAT3D captured INFRARED and RAW Cyclone Imagery over the Indian Ocean from 2012 to 2021. The raw data, sourced from the MOSDAC server, has been meticulously labeled by timestamp and corresponding coordinates on the intensity-time graph of each cyclone directory. Our methodology involves training the CNN model on this extensive dataset and subsequently validating its efficacy by predicting satellite images of cyclone Michaung obtained from the MODIS satellite. The integration of both INSAT3D and MODIS imagery enhances the robustness of our model, providing a more comprehensive understanding of cyclone dynamics. Results indicate a high level of accuracy in cyclone identification, showcasing the potential of CNN techniques in satellite image analysis for cyclone monitoring. The approach not only contributes to the field of meteorology but also establishes a framework for the utilization of diverse satellite datasets in training neural networks for real-time cyclone detection and tracking. This research represents a significant step toward leveraging advanced machine learning techniques to enhance the efficiency and accuracy of tropical cyclone monitoring systems, ultimately aiding in early warning and disaster preparedness efforts.

Development of an Intrusion Detection System Leveraging Deep Learning Model Classification

The implementation of Deep Learning in development of models to act as interventions for addressing the continuously evolving spate of cybersecurity issues has become a noteworthy paradigm. This occurs since cyber attacks could be modeled or represented in terms of data records which can serve as bases for developing intrusion detection systems. This paper proposes an intrusion detection system that leverages deep learning techniques for attack classification. Two deep learning models were developed, a Deep Neural Network (DNN) with ReLU activation as well as Tabular model using the fastai deep learning library. The NSL-KDD benchmark dataset was imported and preprocessed for each model development. After evaluating the models, the fastai model with accuracy of 84 percent surpassed the other DNN with accuracy of 79 percent on NSL-KDD test data.

Advanced Techniques for Biometric Authentication: Leveraging Deep Learning and Explainable AI

Advanced Techniques for Biometric Authentication: Leveraging Deep Learning and Explainable AI

Development of a Low-Cost Livestock Sorting Information Management System Leveraging Deep Learning, AI, and IoT Technologies

The implementation of effective livestock management methods is crucial to optimize agricultural operations. However, conventional livestock sorting and data management approaches encounter several obstacles regarding precision, labor requirements, and financial implications. The process exhibits inefficiency, increased labor costs, and an elevated risk of zoonotic infections. Housing livestock in extensive groups might intensify the transmission of diseases and complicate the surveillance and management of diseased animals. This study attempted to develop a Low-Cost Livestock Sorting Information Management System (LC-LSIMS) using a dataset enriched with crucial metrics and curated images collected over 24 months with the Internet of Things (IoT) and Artificial Intelligence (AI). The design of edge-cloud computing facilitates the redistribution of computational resources, leading to enhanced computational speed. The LC-LSIMS would have a predictive module to assist agricultural practitioners in safeguarding their crops during flood occurrences. This module will empower farmers to proactively anticipate natural phenomena, including floods, during intense rainfall. LC-LSIMS presents a multi-level design plan that facilitates attaining the specified goals. The findings obtained from the execution of the implemented system demonstrate a sorting accuracy of 91.47%, computational speed of 27.42 frames per second (fps), labor cost reduction of 50.84%, production efficiency improvement of 29.59%, and an average reduction in data input errors of 37.59%.

Optimizing Wind Energy Production: Leveraging Deep Learning Models Informed with On-Site Data and Assessing Scalability through HPC

Optimizing Wind Energy Production: Leveraging Deep Learning Models Informed with On-Site Data and Assessing Scalability through HPC

Enhancement of Infrared Imagery through Low‐Light Image Guidance Leveraging Deep Learning Techniques

Enhancement of Infrared Imagery through Low‐Light Image Guidance Leveraging Deep Learning Techniques

AstroSer: Leveraging Deep Learning for Efficient Content-based Retrieval in Massive Solar-observation Images

Rapid and proficient data retrieval is an essential component of modern astronomical research. In this paper, we address the challenge of retrieving astronomical image content by leveraging state-of-the-art deep learning techniques. We have designed a retrieval model, HybridVR, that integrates the capabilities of the deep learning models ResNet50 and VGG16 and have used it to extract key features of solar activity and solar environmental characteristics from observed images. This model enables efficient image matching and allows for content-based image retrieval (CBIR). Experimental results demonstrate that the model can achieve up to 98% similarity during CBIR while exhibiting adaptability and scalability. Our work has implications for astronomical research, data management, and education, and it can contribute to optimizing the utilization of astronomical image data. It also serves as a useful example of the application of deep learning technology in the field of astronomy.

Cancer Diagnosis through Contour Visualization of Gene Expression Leveraging Deep Learning Techniques.

Prompt diagnostics and appropriate cancer therapy necessitate the use of gene expression databases. The integration of analytical methods can enhance detection precision by capturing intricate patterns and subtle connections in the data. This study proposes a diagnostic-integrated approach combining Empirical Bayes Harmonization (EBS), Jensen-Shannon Divergence (JSD), deep learning, and contour mathematics for cancer detection using gene expression data. EBS preprocesses the gene expression data, while JSD measures the distributional differences between cancerous and non-cancerous samples, providing invaluable insights into gene expression patterns. Deep learning (DL) models are employed for automatic deep feature extraction and to discern complex patterns from the data. Contour mathematics is applied to visualize decision boundaries and regions in the high-dimensional feature space. JSD imparts significant information to the deep learning model, directing it to concentrate on pertinent features associated with cancerous samples. Contour visualization elucidates the model's decision-making process, bolstering interpretability. The amalgamation of JSD, deep learning, and contour mathematics in gene expression dataset analysis diagnostics presents a promising pathway for precise cancer detection. This method taps into the prowess of deep learning for feature extraction while employing JSD to pinpoint distributional differences and contour mathematics for visual elucidation. The outcomes underscore its potential as a formidable instrument for cancer detection, furnishing crucial insights for timely diagnostics and tailor-made treatment strategies.

Abstract 14595: Leveraging Deep Learning to Elucidate Subclinical Aortic Stenosis Risk

Background and Aims: Aortic stenosis (AS) is associated with a high global burden of morbidity. Current guidelines do not fully define normal aortic valve (AV) function or mild AS. We aimed to study normal variation in AV function and address current gaps in knowledge. Methods: We developed a deep learning model to study velocity-encoded magnetic resonance imaging (MRI) in 47,223 UK Biobank participants. We calculated aortic valve area (AVA), peak velocity (PV), and mean gradient (MG). Results: In healthy individuals, we found an annual decrement of 0.03 cm2 in AVA. Normal variation in AV function was predictive of future AV surgery (HR 2.4 per SD increase in PV, P2.6E-110), heart failure (HR 1.23 per SD, P=2.4E-07), and coronary disease (HR 1.23 per SD, P=2.0E-14). Risk was still elevated when excluding participants with moderate-to-severe AS. We determined thresholds for abnormal AV function as being >95% for all age groups in 31,909 healthy people. This established the boundary between normal variation and mild AS as any of: PV >1.65m/s, MG >4.9mmHg, and AVA <2.1cm2 (men) or <1.7cms2 (women). A total of 3,288 participants (7.0%) met the definition of mild AS which was associated with high risk for AV surgery (HR 57.2, P=4.4E-29). Over a mean 3.1±2.0 years of follow-up, the incidence of AV surgery was 0.02% in those with normal AV function and nearly 20% among participants with MG ≥10mmHg. Greater levels of ApoB, triglycerides, Lp(a), and inflammatory markers were associated with higher AV gradients. Conclusion: In this study, we report the largest analysis of AV function and cardiovascular disease in the general population with MRI using deep learning. We propose thresholds to define the lower bounds of mild AS and demonstrate a step change in cardiovascular risk between normal AV function and mild AS. Atherogenic and inflammatory biomarkers were associated with increased AV gradients. These findings may inform earlier surveillance and efforts to prevent complications of AS.