Parkpoom Chaisiriprasert
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Parkpoom Chaisiriprasert has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Karn Yongsiriwit
College of Digital Innovation Technology, Rangsit University, Pathumthani 12000, Thailand
4 shared publications - Nattapat Patchsuwan
College of Digital Innovation Technology, Rangsit University, 52/347 Muang-Ake Phaholyothin Road, Lak-Hok, Muang, Pathum Thani 12000, Thailand
3 shared publications - Apicha Deearom
Affiliation College of Digital Innovation Technology, Rangsit University, Pathum Thani, Thailand
2 shared publications - Khachonkit Chuiad
College of Digital Innovation Technology, Rangsit University, Lak Hok, Pathum Thani, Thailand
2 shared publications - Thannob Aribarg
College of Digital Innovation Technology, Rangsit University, Pathumthani 12000, Thailand
2 shared publications Apiporn Simapornchai
Asian Institute of Technology
1 shared publication
Latest Publications
- Open Access
- 9 June 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (6) , e0344942
A deep learning-based fusion framework for robust fine-grained classification of sea turtles in support of marine biodiversity
Accurate classification of sea turtle species is crucial for ecological monitoring and conservation, yet traditional visual classification methods remain limited by underwater imaging challenges such as occlusions, poor lighting, and background noise. To address these limitations, we propose an enhanced deep learning-based classification framework that integrates both color and structural features to improve the robustness of species recognition in complex marine environments. Building upon the ResNet-50 backbone, we introduce a four-channel input tensor comprising RGB data and Sobel-filtered edge maps, capturing both semantic and morphological information. Two novel fusion modules, LiteAFNet and AlphaBlendNet, are designed to integrate these features effectively. LiteAFNet leverages a lightweight attention mechanism to highlight discriminative regions, while AlphaBlendNet adaptively balances RGB and edge cues based on spatial context. Experimental results demonstrate significant improvements in classification performance across all evaluation metrics. Specifically, AlphaBlendNet achieves the highest precision (0.84), recall (0.88), F1-score (0.86), and mean average precision (mAP) of 87.2%, outperforming both the baseline fusion and LiteAFNet configurations. These results indicate that integrating color histograms with structural edge features enhances the model’s ability to distinguish between species with similar visual traits. This framework offers a scalable, accurate, and automated solution for underwater species classification and holds potential for broader application in marine biodiversity monitoring.
- 9 June 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (6) , e0344942
Open AccessA deep learning-based fusion framework for robust fine-grained classification of sea turtles in support of marine biodiversity
Accurate classification of sea turtle species is crucial for ecological monitoring and conservation, yet traditional visual classification methods remain limited by underwater imaging challenges such as occlusions, poor lighting, and background noise. To address these limitations, we propose an enhanced deep learning-based classification framework that integrates both color and structural features to improve the robustness of species recognition in complex marine environments. Building upon the ResNet-50 backbone, we introduce a four-channel input tensor comprising RGB data and Sobel-filtered edge maps, capturing both semantic and morphological information. Two novel fusion modules, LiteAFNet and AlphaBlendNet, are designed to integrate these features effectively. LiteAFNet leverages a lightweight attention mechanism to highlight discriminative regions, while AlphaBlendNet adaptively balances RGB and edge cues based on spatial context. Experimental results demonstrate significant improvements in classification performance across all evaluation metrics. Specifically, AlphaBlendNet achieves the highest precision (0.84), recall (0.88), F1-score (0.86), and mean average precision (mAP) of 87.2%, outperforming both the baseline fusion and LiteAFNet configurations. These results indicate that integrating color histograms with structural edge features enhances the model’s ability to distinguish between species with similar visual traits. This framework offers a scalable, accurate, and automated solution for underwater species classification and holds potential for broader application in marine biodiversity monitoring.
- Open Access
- 13 May 2026
- Published by MDPI AG in International Journal of Environmental Research and Public Health
- Vol. 23 (5) , 650
Visible Helmet Presence Among Motorcycle Riders and Passengers at a Thai University: A CCTV-Based Cross-Sectional Observational Study
Motorcycle-related injuries remain a major public health concern in Thailand, particularly among adolescents and young adults, yet evidence on actual helmet-related behavior in university settings remains limited. This study assessed visible helmet presence among motorcycle riders and passengers at a Thai university using real-world CCTV observations. A cross-sectional observational study was conducted at a large private university in Thailand using CCTV footage from two campus gates over seven consecutive days during peak commuting periods (07:00–09:00 and 16:00–18:00). Motorcycle-user observation events were coded for visible helmet presence, rider/passenger role, university affiliation, and time of observation. Descriptive statistics, chi-square tests, and 95% confidence intervals were used. A total of 19,363 motorcycle-user observation events were recorded. Overall visible helmet presence was 18.8% (3646/19,363). Visible helmet presence was 21.8% (1956/8985) among student riders, 3.1% (230/7505) among student passengers, 58.6% (1299/2217) among staff riders, and 24.5% (161/656) among staff passengers. Morning observation periods showed higher values than evening periods across most groups. Visible helmet presence was low overall and especially low among student passengers. These findings should be interpreted as event-based estimates rather than person-based prevalence or verified protective helmet use, and they support universities as priority settings for targeted road-safety interventions.
- 13 May 2026
- Published by MDPI AG in International Journal of Environmental Research and Public Health
- Vol. 23 (5) , 650
Open AccessVisible Helmet Presence Among Motorcycle Riders and Passengers at a Thai University: A CCTV-Based Cross-Sectional Observational Study
Motorcycle-related injuries remain a major public health concern in Thailand, particularly among adolescents and young adults, yet evidence on actual helmet-related behavior in university settings remains limited. This study assessed visible helmet presence among motorcycle riders and passengers at a Thai university using real-world CCTV observations. A cross-sectional observational study was conducted at a large private university in Thailand using CCTV footage from two campus gates over seven consecutive days during peak commuting periods (07:00–09:00 and 16:00–18:00). Motorcycle-user observation events were coded for visible helmet presence, rider/passenger role, university affiliation, and time of observation. Descriptive statistics, chi-square tests, and 95% confidence intervals were used. A total of 19,363 motorcycle-user observation events were recorded. Overall visible helmet presence was 18.8% (3646/19,363). Visible helmet presence was 21.8% (1956/8985) among student riders, 3.1% (230/7505) among student passengers, 58.6% (1299/2217) among staff riders, and 24.5% (161/656) among staff passengers. Morning observation periods showed higher values than evening periods across most groups. Visible helmet presence was low overall and especially low among student passengers. These findings should be interpreted as event-based estimates rather than person-based prevalence or verified protective helmet use, and they support universities as priority settings for targeted road-safety interventions.
- 2 January 2026
- Published by Springer Nature
- p. 337-348
A Vision Transformer-Based Model for Optimizing Plant Disease Classification
- 2 January 2026
- Published by Springer Nature
- p. 337-348
A Vision Transformer-Based Model for Optimizing Plant Disease Classification
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 165258-165281
A Signal-Aware Framework for Data Quality Assessment and Missing Data Mitigation in LoRaWAN Environmental IoT Networks
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 165258-165281
Open AccessA Signal-Aware Framework for Data Quality Assessment and Missing Data Mitigation in LoRaWAN Environmental IoT Networks
- Open Access
- 15 December 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (24) , 13160
Context-Aware Visual Emotion Recognition Through Hierarchical Fusion of Facial Micro-Features and Scene Semantics
Visual emotion recognition in unconstrained environments remains challenging, as single-stream deep learning models often fail to capture the localized facial cues and contextual information necessary for accurate classification. This study introduces a hierarchical multi-level feature fusion framework that systematically combines low-level micro-textural features (Local Binary Patterns), mid-level facial cues (Facial Action Units), and high-level scene semantics (Places365) with ResNet-50 global embeddings. Evaluated on the large-scale EmoSet-3.3M dataset, which contains 3.3 million images across eight emotion categories, the framework demonstrates marked performance gains with the best configuration (LBP-FAUs-Places365-ResNet). The proposed framework achieves 74% accuracy and a macro-averaged F1-score of 0.75 under its best configuration (LBP-FAUs-Places365-ResNet), representing a five-percentage-point improvement over the ResNet-50 baseline. The approach excels at distinguishing high-intensity emotions, maintaining efficient inference (2.2 ms per image, 29 M parameters), and analysis confirms that integrating facial muscle activations with scene context enables nuanced emotional differentiation. These results validate that hierarchical feature integration significantly advances robust, human-aligned visual emotion recognition, making it suitable for real-world Human–Computer Interaction (HCI) and affective computing applications.
- 15 December 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (24) , 13160
Open AccessContext-Aware Visual Emotion Recognition Through Hierarchical Fusion of Facial Micro-Features and Scene Semantics
Visual emotion recognition in unconstrained environments remains challenging, as single-stream deep learning models often fail to capture the localized facial cues and contextual information necessary for accurate classification. This study introduces a hierarchical multi-level feature fusion framework that systematically combines low-level micro-textural features (Local Binary Patterns), mid-level facial cues (Facial Action Units), and high-level scene semantics (Places365) with ResNet-50 global embeddings. Evaluated on the large-scale EmoSet-3.3M dataset, which contains 3.3 million images across eight emotion categories, the framework demonstrates marked performance gains with the best configuration (LBP-FAUs-Places365-ResNet). The proposed framework achieves 74% accuracy and a macro-averaged F1-score of 0.75 under its best configuration (LBP-FAUs-Places365-ResNet), representing a five-percentage-point improvement over the ResNet-50 baseline. The approach excels at distinguishing high-intensity emotions, maintaining efficient inference (2.2 ms per image, 29 M parameters), and analysis confirms that integrating facial muscle activations with scene context enables nuanced emotional differentiation. These results validate that hierarchical feature integration significantly advances robust, human-aligned visual emotion recognition, making it suitable for real-world Human–Computer Interaction (HCI) and affective computing applications.
- 23 November 2025
- Published by Springer Nature
- p. 317-327
Cardiac Hypertrophy Detection Using ResNet-50: A Deep Learning Approach
- 23 November 2025
- Published by Springer Nature
- p. 317-327
Cardiac Hypertrophy Detection Using ResNet-50: A Deep Learning Approach
- 23 November 2025
- Published by Springer Nature
- p. 329-339
Analysis of Abnormal Driver Behavior from Gaze Direction and Blinking
- 23 November 2025
- Published by Springer Nature
- p. 329-339
Analysis of Abnormal Driver Behavior from Gaze Direction and Blinking
- Open Access
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 194489-194508
Improving Tire Pattern Recognition Using Parameter-Efficient Fine-Tuning Techniques
Tire-tread classification plays a key role in forensic investigation and public safety. This work introduces a robust, efficient recognition system that integrates Discrete Wavelet Transform (DWT) with Weighted Local Gray-Level on Robust Local Binary Pattern (WLG-RLBP), followed by weighted image fusion, and then feeds the result to YOLO-family detectors (v5/v8/v9). For class-incremental learning without access to old-class images, we adopt a unified label space and Label Distillation with Recall-Guard (LD): a frozen teacher generates pseudo-labels for old classes on new images via a two-stage selection (a high-confidence pass, then recall-oriented selection with per-class quotas and IoU de-duplication) so that supervision for old classes is available without storing legacy data. We also embed parameter-efficient fine-tuning (PEFT) in the same objective by freezing the backbone and updating lightweight adapters in the neck/head, reducing training time and memory. Evaluation spans two datasets (a public tire-impression set and a researcher-curated community set). As a reference baseline, full training on the public set yields mAP 87.73% with YOLOv8. Under LD + PEFT, using the same protocol across YOLOv5/8/9, the student models maintain high accuracy (mAP up to 94.1%) while significantly reducing time and memory. Grad-CAM visualizations show that predictions rely on ribs/blocks and sipes rather than background clutter, supporting the method’s forensic credibility.
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 194489-194508
Open AccessImproving Tire Pattern Recognition Using Parameter-Efficient Fine-Tuning Techniques
Tire-tread classification plays a key role in forensic investigation and public safety. This work introduces a robust, efficient recognition system that integrates Discrete Wavelet Transform (DWT) with Weighted Local Gray-Level on Robust Local Binary Pattern (WLG-RLBP), followed by weighted image fusion, and then feeds the result to YOLO-family detectors (v5/v8/v9). For class-incremental learning without access to old-class images, we adopt a unified label space and Label Distillation with Recall-Guard (LD): a frozen teacher generates pseudo-labels for old classes on new images via a two-stage selection (a high-confidence pass, then recall-oriented selection with per-class quotas and IoU de-duplication) so that supervision for old classes is available without storing legacy data. We also embed parameter-efficient fine-tuning (PEFT) in the same objective by freezing the backbone and updating lightweight adapters in the neck/head, reducing training time and memory. Evaluation spans two datasets (a public tire-impression set and a researcher-curated community set). As a reference baseline, full training on the public set yields mAP 87.73% with YOLOv8. Under LD + PEFT, using the same protocol across YOLOv5/8/9, the student models maintain high accuracy (mAP up to 94.1%) while significantly reducing time and memory. Grad-CAM visualizations show that predictions rely on ribs/blocks and sipes rather than background clutter, supporting the method’s forensic credibility.
- Open Access
- 12 November 2025
- Published by MDPI AG in Sustainability
- Vol. 17 (22) , 10091
Toward Sustainable Water Resource Management Using a DWT-NARX Model for Reservoir Inflow and Discharge Forecasting in the Chao Phraya River Basin, Thailand
The 2011 Great Flood in Thailand exposed critical deficiencies in water management across the Chao Phraya River Basin, particularly in controlling inflows and discharges from major reservoirs such as Sirikit and Bhumibol. Inadequate rainfall monitoring at the Nakhon Sawan station further intensified the disaster’s impact. As climate change continues to amplify extreme weather events, this study aims to improve flood forecasting accuracy and promote sustainable water resource management aligned with the Sustainable Development Goals (SDGs 6, 11, and 13). Advanced climate data from the Coupled Model Intercomparison Project Phase 5 (CMIP5) were spatially refined and integrated with hydrological models to enhance regional accuracy. The Discrete Wavelet Transform (DWT) was applied for feature extraction to capture hydrological variability, while the Nonlinear Autoregressive Model with Exogenous Factors (NARX) was employed to model complex temporal relationships. A multi-model ensemble framework was developed to merge climate forecasts with real-time hydrological data. Results demonstrate significant model performance improvements, with DWT-NARX achieving 55–98% lower prediction errors (RMSE) compared to baseline methods and correlation coefficients exceeding 0.91 across all forecasting scenarios. Marked seasonal variations emerge, with higher inflows during wet periods and reduced inflows during dry seasons. Under RCP8.5 climate scenarios, wet-season inflows are projected to increase by 15.8–17.4% by 2099, while dry-season flows may decline by up to 33.5%, potentially challenging future water availability and flood control operations. These findings highlight the need for adaptive and sustainable water management strategies to enhance climate resilience and advance SDG targets on water security, disaster risk reduction, and climate adaptation.
- 12 November 2025
- Published by MDPI AG in Sustainability
- Vol. 17 (22) , 10091
Open AccessToward Sustainable Water Resource Management Using a DWT-NARX Model for Reservoir Inflow and Discharge Forecasting in the Chao Phraya River Basin, Thailand
The 2011 Great Flood in Thailand exposed critical deficiencies in water management across the Chao Phraya River Basin, particularly in controlling inflows and discharges from major reservoirs such as Sirikit and Bhumibol. Inadequate rainfall monitoring at the Nakhon Sawan station further intensified the disaster’s impact. As climate change continues to amplify extreme weather events, this study aims to improve flood forecasting accuracy and promote sustainable water resource management aligned with the Sustainable Development Goals (SDGs 6, 11, and 13). Advanced climate data from the Coupled Model Intercomparison Project Phase 5 (CMIP5) were spatially refined and integrated with hydrological models to enhance regional accuracy. The Discrete Wavelet Transform (DWT) was applied for feature extraction to capture hydrological variability, while the Nonlinear Autoregressive Model with Exogenous Factors (NARX) was employed to model complex temporal relationships. A multi-model ensemble framework was developed to merge climate forecasts with real-time hydrological data. Results demonstrate significant model performance improvements, with DWT-NARX achieving 55–98% lower prediction errors (RMSE) compared to baseline methods and correlation coefficients exceeding 0.91 across all forecasting scenarios. Marked seasonal variations emerge, with higher inflows during wet periods and reduced inflows during dry seasons. Under RCP8.5 climate scenarios, wet-season inflows are projected to increase by 15.8–17.4% by 2099, while dry-season flows may decline by up to 33.5%, potentially challenging future water availability and flood control operations. These findings highlight the need for adaptive and sustainable water management strategies to enhance climate resilience and advance SDG targets on water security, disaster risk reduction, and climate adaptation.
- Open Access
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
LCAT: A Lightweight Color-Aware Transformer with Hierarchical Attention for Leaf Disease Classification in Precision Agriculture
Leaf disease classification plays a critical role in ensuring healthy crop production and preventing agricultural losses. This research proposes a novel deep learning-based method for classifying leaf diseases using a dataset of 21,733 images across six distinct disease categories. We introduce the Lightweight Color-Aware Transformer (LCAT), a model designed to achieve high accuracy and computational efficiency for practical deployment, particularly in environments with limited hardware resources. LCAT uses a special attention system that combines color-focused attention with attention to different sizes of distortions, helping the model to accurately identify disease-related changes that usually appear as slight color changes and small deformations on leaves. A key contribution of LCAT lies in its use of smaller patch sizes to extract features in high-resolution regions while maintaining shallow depth, which significantly reduces model complexity. By focusing attention on a small, highly relevant region of leaf images, LCAT maintains strong classification performance without the computational overhead typically encountered in deeper transformer architectures. The experimental results show that LCAT outperforms standard Vision Transformer (ViT) models with a similar number of parameters. In particular, LCAT achieves a mean average precision (mAP) of 0.81 and a classification accuracy of 0.75, compared to ViT’s mAP of 0.75 and an accuracy of 0.68, while using significantly fewer floating-point operations (FLOPs), at 3.33G vs. 17.58G. This work highlights the potential of transformer-based models in plant pathology and provides a lightweight, high-performance alternative suitable for real-time field or mobile applications. Future work includes extending LCAT to support more crop species and plant diseases and integrating it into comprehensive decision support systems for precision agriculture.
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessLCAT: A Lightweight Color-Aware Transformer with Hierarchical Attention for Leaf Disease Classification in Precision Agriculture
Leaf disease classification plays a critical role in ensuring healthy crop production and preventing agricultural losses. This research proposes a novel deep learning-based method for classifying leaf diseases using a dataset of 21,733 images across six distinct disease categories. We introduce the Lightweight Color-Aware Transformer (LCAT), a model designed to achieve high accuracy and computational efficiency for practical deployment, particularly in environments with limited hardware resources. LCAT uses a special attention system that combines color-focused attention with attention to different sizes of distortions, helping the model to accurately identify disease-related changes that usually appear as slight color changes and small deformations on leaves. A key contribution of LCAT lies in its use of smaller patch sizes to extract features in high-resolution regions while maintaining shallow depth, which significantly reduces model complexity. By focusing attention on a small, highly relevant region of leaf images, LCAT maintains strong classification performance without the computational overhead typically encountered in deeper transformer architectures. The experimental results show that LCAT outperforms standard Vision Transformer (ViT) models with a similar number of parameters. In particular, LCAT achieves a mean average precision (mAP) of 0.81 and a classification accuracy of 0.75, compared to ViT’s mAP of 0.75 and an accuracy of 0.68, while using significantly fewer floating-point operations (FLOPs), at 3.33G vs. 17.58G. This work highlights the potential of transformer-based models in plant pathology and provides a lightweight, high-performance alternative suitable for real-time field or mobile applications. Future work includes extending LCAT to support more crop species and plant diseases and integrating it into comprehensive decision support systems for precision agriculture.