KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 57

Some of the metrics are blocked by your
consent settings
Comparative Analysis Of Convolutional Neural Networks, Long Short-Term Memory Networks, and Bert For Text-Based Emotion Classification
(2026-06-01)
Visutsak, Porawat
;
Tongbai, Tanajak
;
Ongrungruaeng, Duongduen
;
Phongwuttisak, Nuttiruj
;
Wiriya, Surapong
This paper presents a detailed experimental study and comparative analysis of three popular deep learning architectures (CNN, LSTM, and BERT) for emotion classification in written messages. Using a publicly available dataset of six unique emotional states (sadness, joy, love, anger, fear, and surprise), an effective ablation study was conducted to determine optimal architectural configurations, including a sequence length of 66 tokens and an embedding size of 200. To validate the results of a comparative analysis of model performance, a bootstrap technique (30 trials) and the Wilcoxon Signed-Rank test were used to eliminate potential bias. As shown by experiments, the tuned BERT architecture (with a learning rate of 2e-5) produced the most accurate and reliable result of 93.50% in classifying emotional states from texts. Moreover, with an appropriate sequence length configuration, the LSTM network (89.92%) significantly outperformed the CNN (89.65%), confirming the need to account for long-range dependencies in emotion classification. Overall, the research results show the key importance of hyperparameter tuning and the ability to handle complex information for emotion identification.
Some of the metrics are blocked by your
consent settings
Cross-Sensor and Cross-Population Generalization of Deep Learning Models for Digital Mammography: A Controlled Four-Country Benchmark of Five Backbone Architectures with Statistical Significance Testing
(2026-06-01)
Gabbualoy, Somprasonk
;
Phasukkit, Pattarapong
;
Tungjitkusolmun, Supan
Background/Objectives: Deep learning models for digital mammography sensor data are increasingly deployed across hospitals using different X-ray detector technologies and patient populations. Whether models trained on one sensor platform and population maintain accuracy when transferred to another has not been tested for the latest generation of mammography-specific foundation models under one controlled protocol. Methods: We fine-tuned five backbone architectures (ResNet-50, DINOv2-B14, Rad-DINO, Mammo-CLIP B5, and Mammo-FM) on CBIS-DDSM (film-digitized, USA, n = 714 validation) with three seeds, ablated a density-aware focal loss across three auxiliary weights, and evaluated transfer to three external sensor cohorts: CMMD (full-field digital, China, n = 1032), DMID (mixed digital, India, n = 509), and MIAS (film-digitized, UK, n = 322). Significance used paired DeLong z-tests with Benjamini–Hochberg FDR correction; temperature scaling tested post hoc recalibration at all transfer targets. Results: Within this single-source three-seed evaluation, ResNet-50 outperformed all four foundation models on CBIS-DDSM (AUC 0.867 vs. 0.847, 0.846, 0.813, and 0.703; all gaps p_adj < 0.05). The density-aware focal loss degraded both AUC and calibration at every weight tested. At transfer, every model lost 0.165 to 0.320 AUC points relative to in-distribution performance, with sensitivity at 95% specificity collapsing from 0.31 to 0.47 in-distribution to 0.11 to 0.22 across the three external targets. A per-seed Stouffer meta-analysis confirms that Mammo-CLIP B5 and Mammo-FM significantly outperformed ResNet-50 on DMID and Mammo-CLIP on CMMD, after BH-FDR; MIAS comparisons remained directional only. In the extremely dense subgroup (BI-RADS D4), Mammo-FM reached AUC 0.870 versus ResNet-50 at 0.842, a directional observation whose 95% CIs overlap heavily at the n = 140 sample size and which we do not interpret as a statistically supported advantage. Conclusions: In this single training-source, three-seed protocol, mammography-specific pretraining did not deliver the in-distribution AUC premium reported in the originating papers, and no architecture reached a level at which transfer deployment without local validation would be defensible. We frame these as observations specific to the present protocol rather than as broader conclusions about foundation models for mammography classification. The findings argue for sensor-stratified and population-stratified external validation and for local recalibration as practical prerequisites before clinical use. Code and weights are released under MIT license.
Some of the metrics are blocked by your
consent settings
Smart Diagnostics: Hierarchical Deep Learning of Acoustic Emission Signals for Early Crack Detection in Zirconia Dental Structures
(2026-05-01)
Tuntiwong, Kuson
;
Wangman, Rangsinee
;
Kanchanatawewat, Kanchana
;
Anucul, Boonjira
;
Sritart, Hiranya
Monolithic zirconia restorations are frequently affected by the unnoticed growth of subcritical cracks, a failure process that is not captured by traditional imaging methods like radiographs and ultrasounds in sophisticated dental architectures. To address this evaluative inadequacy, this research introduces a hierarchical deep learning framework for microcrack detection and spatial localization. We promote a hierarchical deep learning system that integrates Acoustic Emission (AE) detection alongside signal processing. Raw AE signals utilized during dynamic loading are enhanced via Kalman filtering and Continuous Wavelet Transform (CWT) to construct high-fidelity time–frequency scalograms. The diagnostic pipeline operates in two stages: first, a hybrid CNN–BiGRU network with temporal attention fulfills zirconia component-level classification; second, a ResNet-18 backbone integrated with Bidirectional LSTM and Multi-Head Attention precisely localizes defects across five anatomical crown regions. This hierarchical design effectively captures the non-stationary, transient nature of fracture-induced stress waves. The framework achieved an F1-score of 99.00% and an AUC of 0.994, significantly outperforming conventional convolutional networks. By enabling predictive maintenance through early, non-invasive damage localization, this study demonstrates a promising laboratory framework for AE-based crack detection in zirconia dental structures and prosthetics and toward enhanced clinical reliability in digital dentistry.
Some of the metrics are blocked by your
consent settings
Day–Night All-Sky Scene Classification with an Attention-Enhanced EfficientNet
(2026-02-01)
Boonpook, Wuttichai
;
Torteeka, Peerapong
;
Torsri, Kritanai
;
Kamthonkiat, Daroonwan
;
Tan, Yumin
All-sky cameras provide continuous hemispherical observations essential for atmospheric monitoring and observatory operations; however, automated classification of sky conditions in tropical environments remains challenging due to strong illumination variability, atmospheric scattering, and overlapping thin-cloud structures. This study proposes EfficientNet-Attention-SPP Multi-scale Network (EASMNet), a physics-aware deep learning framework for robust all-sky scene classification using hemispherical imagery acquired at the Thai National Observatory. The proposed architecture integrates Squeeze-and-Excitation (SE) blocks for radiometric channel stabilization, the Convolutional Block Attention Module (CBAM) for spatial–semantic refinement, and Spatial Pyramid Pooling (SPP) for hemispherical multi-scale context aggregation within a fully fine-tuned EfficientNetB7 backbone, forming a domain-aware atmospheric representation framework. A large-scale dataset comprising 122,660 RGB images across 13 day–night sky-scene categories was curated, capturing diverse tropical atmospheric conditions including humidity, haze, illumination transitions, and sensor noise. Extensive experimental evaluations demonstrate that the EASMNet achieves 93% overall accuracy, outperforming representative convolutional (VGG16, ResNet50, DenseNet121) and transformer-based architectures (Swin Transformer, Vision Transformer). Ablation analyses confirm the complementary contributions of hierarchical attention and multi-scale aggregation, while class-wise evaluation yields F1-scores exceeding 0.95 for visually distinctive categories such as Day Humid, Night Clear Sky, and Night Noise. Residual errors are primarily confined to physically transitional and low-contrast atmospheric regimes. These results validate the EASMNet as a reliable, interpretable, and computationally feasible framework for real-time observatory dome automation, astronomical scheduling, and continuous atmospheric monitoring, and provide a scalable foundation for autonomous sky-observation systems deployable across diverse climatic regions.
Some of the metrics are blocked by your
consent settings
IoMT–Fog–Cloud-based AI frameworks for chronic disease diagnosis: updated comparative analysis with recent AI-IoMT models (2020–2025)
(2026-01-01)
Locharoenrat, Kitsakorn
Chronic diseases such as diabetes and cardiovascular disease require frequent monitoring and timely clinical feedback to prevent complications. Internet of Medical Things (IoMT) systems increasingly combine near-patient sensing with Fog and Cloud computing so that time-critical preprocessing and inference can run close to the patient while compute-intensive training and population-level analytics remain in the Cloud. This review synthesizes primary studies published between 2020 and 2025 that implement AI-enabled IoMT, with an emphasis on systems that report both diagnostic performance and network quality-of-service (QoS). Following PRISMA 2020, we screened database records and included 14 primary studies; we focus the joint performance–QoS synthesis on six IoMT–Fog–Cloud frameworks for diabetes and cardiovascular disease and compare them with two recent multi-disease AI-IoMT models (DACL and TasLA). Diabetes-oriented implementations commonly report accuracy around 95%–96% using explainable or ensemble deep learning, whereas some cardiovascular frameworks report >99% accuracy in controlled settings; we therefore discuss plausible sources of optimistic performance, including small datasets, class imbalance, curated benchmarks, and potential leakage/overfitting in simulation-based evaluations. Across IoMT–Fog–Cloud studies, placing preprocessing and/or inference at the Fog layer repeatedly reduces end-to-end latency for streaming biosignals, but multi-Fog provisioning can increase energy and power demands. To support more reproducible comparisons, we organize 14 extracted metrics into (i) diagnostic performance (accuracy, precision, recall, F1-score, sensitivity, specificity) and (ii) system/network QoS (latency, jitter, throughput, bandwidth utilization, processing/execution time, network usage, energy consumption, power consumption), and we translate the evidence into study-linked design recommendations for future deployments.
Some of the metrics are blocked by your
consent settings
Unlocking the Future of Orthopaedic Imaging: A Comprehensive Update on the Role and Benefits of The Medical Open Network for AI (MONAI)
(2025-12-31)
Angthong, Chayanin
;
Angthong, Wirana
;
Pongsakonpruttikul, Napat
BACKGROUND: Deep learning (DL) has revolutionized orthopaedic imaging, transitioning from traditional radiomics-based analysis to powerful, data-driven diagnostic and prognostic models. However, a persistent lack of methodological standardization has limited clinical translation. The Medical Open Network for AI (MONAI), an open-source PyTorch-based framework, addresses this gap by providing domain-specific tools optimized for medical imaging. This review evaluates MONAI's role and benefits in orthopaedics across diagnosis, treatment planning, and outcomes prediction. MATERIAL AND METHODS: A comprehensive literature synthesis was conducted, examining studies utilizing MONAI for musculoskeletal imaging. We assessed technical attributes including architecture, data handling, loss functions, and multimodal integration and their applications in fracture detection, disease grading, surgical planning, and prognostic modeling. RESULTS: MONAI demonstrated superior efficiency in handling 3D/4D orthopaedic imaging data and managing class imbalance using specialized medical loss functions (e.g., Dice and Tversky). Diagnostic models achieved near-expert accuracy in fracture detection and quantitative osteoarthritis grading, providing explainable, and reproducible outputs. MONAI enabled automated, high-fidelity 3D reconstruction for personalized implant design and 3D printing integration. Prognostically, it outperformed surgeons in predicting arthroplasty complications, revealing latent imaging biomarkers. The framework's evolution into MONAI Multimodal - with agentic AI and radiomics integration - enhanced personalized, multimodal risk assessment. CONCLUSIONS: 1. MONAI establishes a standardized, transparent infrastructure that accelerates orthopaedic AI research and clinical translation. Its integration of domain-optimized architectures, multimodal data fusion, and explainable AI tools enables accurate diagnosis, individualized surgical planning, and reliable outcome prediction. 2. Adoption of MONAI-based pipelines is strongly recommended to promote reproducibility, regulatory readiness, and clinician trust in next generation of precision orthopaedic care.
Some of the metrics are blocked by your
consent settings
An Extension of Laor Weight Initialization for Deep Time-Series Forecasting: Evidence from Thai Equity Risk Prediction
(2025-09-01)
Petchpol, Katsamapol
;
Boongasame, Laor
This study presents a gradient-informed proxy initialization framework designed to improve training efficiency and predictive performance in deep learning models for time-series forecasting. The method extends the Laor Initialization approach by introducing backward gradient norm clustering as a selection criterion for input-layer weights, evaluated through a lightweight, architecture-agnostic proxy model. Only the numerical input layer adopts the selected initialization, while internal components retain standard schemes such as Xavier, Kaiming, or Orthogonal, maintaining compatibility and reducing overhead. The framework is evaluated on a real-world financial forecasting task: identifying high-risk equities from the Thai Market Surveillance Measure List, a domain characterized by label imbalance, non-stationarity, and limited data volume. Experiments across five architectures, including Transformer, ConvTran, and MMAGRU-FCN, show that the proposed strategy improves convergence speed and classification accuracy, particularly in deeper and hybrid models. Results in recurrent-based models are competitive but less pronounced. These findings support the method’s practical utility and generalizability for forecasting tasks under real-world constraints.
Some of the metrics are blocked by your
consent settings
Revealing the Power of Deep Learning in Quality Assessment of Mango and Mangosteen Purée Using NIR Spectral Data
(2025-09-01)
Pornchaloempong, Pimpen
;
Sharma, Sneha
;
Phanomsophon, Thitima
;
Sirisomboon, Panmanas
;
Lapcharoensuk, Ravipat
The quality control of fruit purée products such as mango and mangosteen is crucial for maintaining consumer satisfaction and meeting industry standards. Traditional destructive techniques for assessing key quality parameters like the soluble solid content (SSC) and titratable acidity (TA) are labor-intensive and time-consuming; prompting the need for rapid, nondestructive alternatives. This study investigated the use of deep learning (DL) models including Simple-CNN, AlexNet, EfficientNetB0, MobileNetV2, and ResNeXt for predicting SSC and TA in mango and mangosteen purée and compared their performance with the conventional chemometric method partial least squares regression (PLSR). Spectral data were preprocessed and evaluated using 10-fold cross-validation. For mango purée, the Simple-CNN model achieved the highest predictive accuracy for both SSC (coefficient of determination of cross-validation ((Formula presented.)) = 0.914, root mean square error of cross-validation (RMSE<inf>CV</inf>) = 0.688, the ratio of prediction to deviation of cross-validation (RPD<inf>CV</inf>) = 3.367) and TA ((Formula presented.) = 0.762, RMSE<inf>CV</inf> = 0.037, RPD<inf>CV</inf> = 2.864), demonstrating a statistically significant improvement over PLSR. For the mangosteen purée, AlexNet exhibited the best SSC prediction performance ((Formula presented.) = 0.702, RMSE<inf>CV</inf> = 0.471, RPD<inf>CV</inf> = 1.666), though the RPD<inf>CV</inf> values (<2.0) indicated limited applicability for precise quantification. TA prediction in mangosteen purée showed low variance in the reference values (standard deviation (SD) = 0.048), which may have restricted model performance. These results highlight the potential of DL for improving NIR-based quality evaluation of fruit purée, while also pointing to the need for further refinement to ensure interpretability, robustness, and practical deployment in industrial quality control.
Some of the metrics are blocked by your
consent settings
Automatic Fish Classification Using Lanczos Resampling and Deep Learning
(2025-09-01)
Kuswantori, Ari
;
Suesut, Taweepol
;
Suthanupaphwut, Worapanya
;
Tangsrirat, Worapong
;
Nunak, Navaphattra
The development of automation in the fish industry, a vital sector of the food industry, is a highly relevant and essential topic. This development is essential for boosting output and mitigating the risk of future food shortages brought on by the world’s population expansion. Automatic fish classification using computer vision has been widely developed in fish industry automation, and a lot of research on that topic has been published. However, while some research has produced promising results using complex methods, others have applied simpler approaches with less satisfactory outcomes. This study suggests a straightforward but efficient technique for differentiating between fish species by concentrating on their main characteristics, such as body form and scale patterns. To effectively support these image capturing properties, the Lanczos re-sampling technique is used in this study. Additionally, our basic deep learning model can correctly learn and identify fish species thanks to a fish picture categorization engine created using Google Teachable Machine. Utilizing the Fish-Pak dataset, a popular fish picture dataset frequently used in studies on fish species classification, the suggested approach successfully overcomes the difficulty and attains a high accuracy rate of 97.16%.
Some of the metrics are blocked by your
consent settings
Laor Initialization: A New Weight Initialization Method for the Backpropagation of Deep Learning
(2025-07-01)
Boongasame, Laor
;
Muangprathub, Jirapond
;
Thammarak, Karanrat
This paper presents Laor Initialization, an innovative weight initialization technique for deep neural networks that utilizes forward-pass error feedback in conjunction with k-means clustering to optimize the initial weights. In contrast to traditional methods, Laor adopts a data-driven approach that enhances convergence’s stability and efficiency. The method was assessed using various datasets, including a gold price time series, MNIST, and CIFAR-10 across the CNN and LSTM architectures. The results indicate that the Laor Initialization achieved the lowest K-fold cross-validation RMSE (0.00686), surpassing Xavier, He, and Random. Laor demonstrated a high convergence success (final RMSE = 0.00822) and the narrowest interquartile range (IQR), indicating superior stability. Gradient analysis confirmed Laor’s robustness, achieving the lowest coefficients of variation (CV = 0.2230 for MNIST, 0.3448 for CIFAR-10, and 0.5997 for gold price) with zero vanishing layers in the CNNs. Laor achieved a 24% reduction in CPU training time for the Gold price data and the fastest runtime on MNIST (340.69 s), while maintaining efficiency on CIFAR-10 (317.30 s). It performed optimally with a batch size of 32 and a learning rate between 0.001 and 0.01. These findings establish Laor as a robust alternative to conventional methods, suitable for moderately deep architectures. Future research should focus on dynamic variance scaling and adaptive clustering.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results