Object detection algorithms to identify skeletal components in carbonate cores DOI Creative Commons
Harriet L. Dawson, Cédric M. John

Marine and Petroleum Geology, Journal Year: 2024, Volume and Issue: 167, P. 106965 - 106965

Published: June 15, 2024

Identification of constituent grains in carbonate rocks requires specialist experience. A sedimentologist must be able to distinguish between skeletal that change through geological ages, preserved differing alteration stages, and cut random orientations across core sections. Recent studies have demonstrated the effectiveness machine learning classifying lithofacies from thin section, core, seismic images, with faster analysis times reduction natural biases. In this study, we explore application limitations convolutional neural network (CNN) based object detection frameworks identify quantify multiple types within close-up images lithologies. We compiled nearly 400 high-resolution three ODP IODP expeditions. Over 9000 individual components 11 different classes were manually labelled dataset. Using pre-trained weights, a transfer approach was applied evaluate one-stage (YOLO v5) two-stage (Faster R–CNN) detectors under feature extractors (CSP-Darknet53 ResNet50-FPN, respectively). Despite current popularity detectors, our results show Faster R–CNN ResNet50-FPN backbone provides most robust performance, achieving 0.73 mean average precision (mAP). Furthermore, extend by deploying trained model two sites Leg 194 not part training set (ODP Sites 1196 1199), providing performance comparison benchmark human interpretation.

Language: Английский

"Challenges and future in deep learning for sentiment analysis: a comprehensive review and a proposed novel hybrid approach" DOI Creative Commons
Md Shofiqul Islam, Muhammad Nomani Kabir, Ngahzaifa Ab Ghani

et al.

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 57(3)

Published: Feb. 19, 2024

Abstract Social media is used to categorise products or services, but analysing vast comments time-consuming. Researchers use sentiment analysis via natural language processing, evaluating methods and results conventionally through literature reviews assessments. However, our approach diverges by offering a thorough analytical perspective with critical analysis, research findings, identified gaps, limitations, challenges future prospects specific deep learning-based in recent times. Furthermore, we provide in-depth investigation into categorizing prevalent data, pre-processing methods, text representations, learning models, applications. We conduct evaluation of advances architectures, assessing their pros cons. Additionally, offer meticulous methodologies, integrating insights on applied tools, strengths, weaknesses, performance results, detailed feature-based examination. present discussion the challenges, drawbacks, factors contributing successful enhancement accuracy within realm analysis. A comparative article clearly shows that capsule-based RNN approaches give best an 98.02% which CNN RNN-based models. implemented various advanced deep-learning models across four benchmarks identify top performers. introduced innovative CRDC (Capsule Deep Bi structured RNN) model, demonstrated superior compared other methods. Our proposed achieved remarkable different databases: IMDB (88.15%), Toxic (98.28%), CrowdFlower (92.34%), ER (95.48%). Hence, this method holds promise for automated potential deployment.

Language: Английский

Citations

21
Hybrid Optimization of Phase Masks: Integrating Non-Iterative Methods with Simulated Annealing and Validation via Tomographic Measurements DOI Open Access

Z. Li,

Chao Sun,

Haihua Wang

et al.

Symmetry, Journal Year: 2025, Volume and Issue: 17(4), P. 530 - 530

Published: March 31, 2025

The development of holography has facilitated significant advancements across a wide range disciplines. A phase-only spatial light modulator (SLM) plays crucial role in realizing digital holography, typically requiring phase mask as its input. Non-iterative (NI) algorithms are widely used for generation, yet they often fall short delivering precise solutions and lack adaptability complex scenarios. In contrast, the Simulated Annealing (SA) algorithm provides global optimization approach capable addressing these limitations. This study investigates integration NI with SA to enhance generation holography. Furthermore, we examine how adjusting annealing parameters, especially cooling strategy, can significantly improve system performance symmetry. Notably, observe considerable improvement efficiency when non-iterative methods employed generate initial mask. Our method achieves perfect representation symmetry desired fields. efficacy optimized masks is evaluated through optical tomographic measurements using two-dimensional mutually unbiased bases (MUBs), resulting average similarity reaching 0.99. These findings validate effectiveness our methodin optimizing underscore potential high-precision mode recognition analysis.

Language: Английский

Citations

3
Exploring the Association Between Textual Parameters and Psychological and Cognitive Factors DOI Creative Commons
Kadir Uludağ

Psychology Research and Behavior Management, Journal Year: 2024, Volume and Issue: Volume 17, P. 1139 - 1150

Published: March 1, 2024

Textual data analysis has become a popular method for examining complex human behavior in various fields, including psychology, psychiatry, sociology, computer science, mining, forensic sciences, and communication studies. However, identifying the most relevant textual parameters analyzing is still challenge.

Language: Английский

Citations

16
Exploring the role of computer vision in product design and development: a comprehensive review DOI
Lu Yang, Raman Kumar, Rupinder Kaur

et al.

International Journal on Interactive Design and Manufacturing (IJIDeM), Journal Year: 2024, Volume and Issue: 18(6), P. 3633 - 3680

Published: March 14, 2024

Language: Английский

Citations

16
Research on Media Text Translation Based on Information Retrieval DOI Creative Commons
Jiuquan Zhang, Meng Yan

International Journal of e-Collaboration, Journal Year: 2025, Volume and Issue: 21(1), P. 1 - 14

Published: March 5, 2025

This paper focuses on the research of media knowledge text translation based information retrieval, and discusses how retrieval technology can affect optimize process results texts. By combining basic principles its practical application scenarios, this analyzes characteristics texts their special requirements for translation. Through design experiment, effect improving quality, efficiency reaction cost is evaluated. The experimental show that significantly improve accuracy fluency translation, shorten cycle reduce cost. prospect in field will be broader. study not only provide a new perspective method texts, but also contribute to improvement cross-cultural communication dissemination efficiency.

Language: Английский

Citations

0
Enhancing Performance of Beam-Column Joints in Reinforced Concrete Structures Using Carbon Fiber-Reinforced Polymers (CFRP): A Novel Review DOI Creative Commons
Gift Onyinyechi Oloni, Abdulkhalik J. Abdulridha

Hybrid Advances, Journal Year: 2025, Volume and Issue: unknown, P. 100444 - 100444

Published: March 1, 2025

Language: Английский

Citations

0
Leveraging the Internet of Behaviors for Mutual Trust in Digital Ecosystems DOI
Hind Bangui, Barbora Bühnová, Mouzhi Ge

et al.

Published: March 18, 2025

Language: Английский

Citations

0
Exploring the prospects of multimodal large language models for Automated Emotion Recognition in education: Insights from Gemini DOI
Shuzhen Yu, Alexey Androsov, Hanbing Yan

et al.

Computers & Education, Journal Year: 2025, Volume and Issue: unknown, P. 105307 - 105307

Published: March 1, 2025

Language: Английский

Citations

0
Emerging Impact of Parental Internet Addiction on Adolescent Internet Use: A Cross‐Cultural Perspective DOI Creative Commons
Ala Yankouskaya, Raian Ali, Sameha Alshakhsi

et al.

New Directions for Child and Adolescent Development, Journal Year: 2025, Volume and Issue: 2025(1)

Published: Jan. 1, 2025

The escalating global concern about internet addiction (IA) in adolescents has driven the necessity to investigate its predictors and their potential effects on youth development. We used a novel methodological approach facilitate this research assessed IA parents across five countries—GCC countries, Greece, Italy, Turkey, United Kingdom. A total of 1530 participants completed surveys evaluating parental IA, monitoring practices, adolescent symptoms. found striking evidence that involvement nonessential online activities, frequent arguments between children were significant IA. Our data suggest similar sociopsychological mechanisms underlying development various cultural contexts. Contrary earlier assumptions, time spent did not predict suggesting simply regulating screen may be insufficient reduce youth. Instead, tight corresponding symptoms parent indicate need for family‐centered interventions mitigate risks.

Language: Английский

Citations

0
Acoustic-based machine learning approaches for depression detection in Chinese university students DOI Creative Commons

Yange Wei,

Shisen Qin,

Fengyi Liu

et al.

Frontiers in Public Health, Journal Year: 2025, Volume and Issue: 13

Published: May 15, 2025

Background Depression is major global public health problems among university students. Currently, the evaluation and monitoring of depression predominantly depend on subjective self-reported methods. There an urgent necessity to develop objective means identifying depression. Acoustic features, which convey emotional information, have potential enhance objectivity assessments. This study aimed investigate feasibility utilizing acoustic features for automated identification characterization Chinese Methods A cross-sectional was undertaken involving 103 students with controls matched age, gender, education. Participants' voices were recorded using a smartphone as they read neutral texts. analysis feature extraction performed OpenSMILE toolkit, yielding 523 encompassing spectral, glottal, prosodic characteristics. These extracted utilized discriminant between control groups. Pearson correlation analyses conducted evaluate relationship Patient Health Questionnaire-9 (PHQ-9) scores. Five machine learning algorithms including Linear Discriminant Analysis (LDA), Logistic Regression, Support Vector Classification, Naive Bayes, Random Forest used perform classification. For training testing, ten-fold cross-validation employed. Model performance assessed receiver operating characteristic (ROC) curve, area under curve (AUC), precision, accuracy, recall, F1 score. Shapley Additive exPlanations (SHAP) method model interpretation. Results In group, 32 (25 spectral 5 2 glottal features) showed significant alterations compared controls. Further, 27 (10 3 1 significantly correlated severity. Among five algorithms, LDA demonstrated highest classification performance, AUC 0.771. SHAP suggested that Mel-frequency cepstral coefficients (MFCC) contributed most model's efficacy. Conclusions The integration demonstrates high accuracy in distinguishing students, suggesting its utility rapid large-scale screening. MFCC may serve valid campuses.

Language: Английский

Citations

0