Olarik Surinta
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Olarik Surinta has a total of 64 co-authors. Here are the top co-authors by number of shared publications:
- Lambert Schomaker
University of Groningen
11 shared publications - Emmanuel Okafor
Center of Artificial Intelligence for Security, Naif Arab University for Security Sciences, Riyadh, 11452, Saudi Arabia
6 shared publications Marco A. Wiering
4 shared publicationsMarco Wiering
University of Groningen, Netherlands
4 shared publicationsThananchai Khamket
Mahasarakham University
4 shared publicationsEmmanuel Okafor
SDAIA-KFUPM Joint Research Center for Artificial Intelligence King Fahd University of Petroleum and Minerals 31261 Dharan Saudi Arabia
3 shared publications
Latest Publications
- Open Access
Symmetry-Preserving Transformer-Based Shape Prediction and Neural Shape Control for Whole-Body Object Caging with Hyper-Redundant Robots
Coordinating whole-body shape generation, geometric object caging, and collision-aware transportation remains challenging for hyper-redundant robots. This study presents a framework integrating Transformer-based backbone prediction, cage-preserving geometric refinement, and neural shape control for a planar 50-link robot operating among static and moving obstacles. The framework was evaluated in five simulated scenes under idealized kinematic assumptions without explicit actuator dynamics, contact forces, or sensing uncertainty. The full implementation achieved a geometric cage success rate of 100%, with mean cage-center and cage-radial errors of 0.001786 and 0.000062, respectively. However, the mean robot–obstacle and object–obstacle collision rates were 75.22% and 15.87%, and none of the five scenes satisfied the complete task-success criterion. A controlled ablation of soft cage attraction produced nearly unchanged performance when hard cage projection was retained, indicating limited incremental benefit from the soft attraction term under the evaluated conditions. This experiment does not isolate the contribution of symmetry preservation as a whole. A dedicated CPU benchmark measured a mean control-cycle latency of 128.792 ± 2.401 ms, with shape optimization accounting for approximately 51.4% of the mean cycle time. The findings demonstrate consistent maintenance of local enclosure geometry but do not establish collision-free transportation or physically feasible manipulation. Further development should prioritize whole-body feasibility, collision-aware motion planning, and validation across additional training seeds and physically realistic environments.
Open AccessSymmetry-Preserving Transformer-Based Shape Prediction and Neural Shape Control for Whole-Body Object Caging with Hyper-Redundant Robots
Coordinating whole-body shape generation, geometric object caging, and collision-aware transportation remains challenging for hyper-redundant robots. This study presents a framework integrating Transformer-based backbone prediction, cage-preserving geometric refinement, and neural shape control for a planar 50-link robot operating among static and moving obstacles. The framework was evaluated in five simulated scenes under idealized kinematic assumptions without explicit actuator dynamics, contact forces, or sensing uncertainty. The full implementation achieved a geometric cage success rate of 100%, with mean cage-center and cage-radial errors of 0.001786 and 0.000062, respectively. However, the mean robot–obstacle and object–obstacle collision rates were 75.22% and 15.87%, and none of the five scenes satisfied the complete task-success criterion. A controlled ablation of soft cage attraction produced nearly unchanged performance when hard cage projection was retained, indicating limited incremental benefit from the soft attraction term under the evaluated conditions. This experiment does not isolate the contribution of symmetry preservation as a whole. A dedicated CPU benchmark measured a mean control-cycle latency of 128.792 ± 2.401 ms, with shape optimization accounting for approximately 51.4% of the mean cycle time. The findings demonstrate consistent maintenance of local enclosure geometry but do not establish collision-free transportation or physically feasible manipulation. Further development should prioritize whole-body feasibility, collision-aware motion planning, and validation across additional training seeds and physically realistic environments.
- Open Access
- 1 August 2026
- Published by Suranaree University of Technology in Suranaree Journal of Science and Technology
- Vol. 33 (4) , 010440(1-17)
ATTENTION-ASTFF-NET: A CHANNEL-AWARE ADAPTIVE SPATIOTEMPORAL FUSION NETWORK FOR VEHICLE RECOGNITION
Vehicle recognition systems face challenges in simultaneously capturing spatial appearance characteristics and temporal structural patterns. This study proposes Attention-ASTFF-Net, a channel-aware adaptive spatiotemporal fusion network that enhances vehicle type and vehicle make recognition. The proposed framework extracts spatial features with a convolutional backbone and temporal representations with a convolutional 1D-based module. An adaptive fusion mechanism then combines these features to obtain a comprehensive spatiotemporal representation. Channel-wise attention mechanisms, including Efficient Channel Attention (ECA-Net), Gated Channel Transformation (GCT), Squeeze-and-Excitation (SE) Networks, and Selective Kernel Networks (SKNet), are integrated to emphasize important feature channels and improve discriminative capability. Experimental results on the Vehicle Type Image Dataset Version 2 (VTID2) and the Vehicle Make Image Dataset (VMID) show that the proposed network consistently outperforms the baseline ASTFF-Net and other convolutional neural network architectures. Channel-wise attention integration improves localization, feature selectivity, and prediction stability. The proposed network achieves over 99% accuracy on VTID2 compared with approximately 96% achieved by previous methods, and over 97% on VMID compared with 93.11% achieved by existing approaches. Furthermore, Attention-ASTFF-Net with the SE module demonstrates strong computational efficiency at 5.113 giga floating point operations.
- 1 August 2026
- Published by Suranaree University of Technology in Suranaree Journal of Science and Technology
- Vol. 33 (4) , 010440(1-17)
Open AccessATTENTION-ASTFF-NET: A CHANNEL-AWARE ADAPTIVE SPATIOTEMPORAL FUSION NETWORK FOR VEHICLE RECOGNITION
Vehicle recognition systems face challenges in simultaneously capturing spatial appearance characteristics and temporal structural patterns. This study proposes Attention-ASTFF-Net, a channel-aware adaptive spatiotemporal fusion network that enhances vehicle type and vehicle make recognition. The proposed framework extracts spatial features with a convolutional backbone and temporal representations with a convolutional 1D-based module. An adaptive fusion mechanism then combines these features to obtain a comprehensive spatiotemporal representation. Channel-wise attention mechanisms, including Efficient Channel Attention (ECA-Net), Gated Channel Transformation (GCT), Squeeze-and-Excitation (SE) Networks, and Selective Kernel Networks (SKNet), are integrated to emphasize important feature channels and improve discriminative capability. Experimental results on the Vehicle Type Image Dataset Version 2 (VTID2) and the Vehicle Make Image Dataset (VMID) show that the proposed network consistently outperforms the baseline ASTFF-Net and other convolutional neural network architectures. Channel-wise attention integration improves localization, feature selectivity, and prediction stability. The proposed network achieves over 99% accuracy on VTID2 compared with approximately 96% achieved by previous methods, and over 97% on VMID compared with 93.11% achieved by existing approaches. Furthermore, Attention-ASTFF-Net with the SE module demonstrates strong computational efficiency at 5.113 giga floating point operations.
- Open Access
Thai–English Multiscript Text Image (TEMS) Dataset for Multiscript Image-Based Text Recognition
Publicly available datasets containing Thai and English scene text captured under unconstrained real-world conditions remain limited, particularly for multilingual and mixed-script text recognition. To address this gap, the Thai–English Multiscript Text Image (TEMS) Dataset was developed as a publicly available resource for multilingual scene text recognition research. Natural scene photographs containing Thai and English textual content were collected using smartphone cameras from diverse environments, including billboards, commercial storefronts, road signs, menus, packaging, and publication covers. Text regions were manually identified, verified, and extracted to generate cropped text images with corresponding transcription labels. The resulting dataset comprises 5000 text images derived from 1625 natural scene photographs and contains Thai text, English text, mixed Thai–English text, numerals, punctuation marks, and special symbols spanning 161 unique character classes. The dataset exhibits substantial variation in text length, image dimensions, font style, text size, illumination conditions, spatial arrangement, and background complexity. Statistical analysis indicates an average text length of 15.8 characters per image. In addition, image widths range from 45 to 1512 pixels and image heights range from 18 to 275 pixels. The TEMS Dataset provides a publicly available multilingual scene text resource for optical character recognition, scene text recognition, document understanding, computer vision, pattern recognition, and multilingual artificial intelligence research.
Open AccessThai–English Multiscript Text Image (TEMS) Dataset for Multiscript Image-Based Text Recognition
Publicly available datasets containing Thai and English scene text captured under unconstrained real-world conditions remain limited, particularly for multilingual and mixed-script text recognition. To address this gap, the Thai–English Multiscript Text Image (TEMS) Dataset was developed as a publicly available resource for multilingual scene text recognition research. Natural scene photographs containing Thai and English textual content were collected using smartphone cameras from diverse environments, including billboards, commercial storefronts, road signs, menus, packaging, and publication covers. Text regions were manually identified, verified, and extracted to generate cropped text images with corresponding transcription labels. The resulting dataset comprises 5000 text images derived from 1625 natural scene photographs and contains Thai text, English text, mixed Thai–English text, numerals, punctuation marks, and special symbols spanning 161 unique character classes. The dataset exhibits substantial variation in text length, image dimensions, font style, text size, illumination conditions, spatial arrangement, and background complexity. Statistical analysis indicates an average text length of 15.8 characters per image. In addition, image widths range from 45 to 1512 pixels and image heights range from 18 to 275 pixels. The TEMS Dataset provides a publicly available multilingual scene text resource for optical character recognition, scene text recognition, document understanding, computer vision, pattern recognition, and multilingual artificial intelligence research.
- Open Access
- 1 September 2025
- Published by Learning Gate in Edelweiss Applied Science and Technology
- Vol. 9 (9) , 486-503
Algorithmic mediation and virtual engagement in smart tourism: An expectation–confirmation model of trust and travel intention among Gen Z tourists in Thailand
Tourism in the digital era is increasingly influenced by algorithmic recommendation systems and immersive online interactions. This study examines how algorithmic mediation and virtual engagement jointly affect Chinese Generation Z tourists' trust and travel intentions in Thailand, a leading destination for smart tourism. Guided by the expectation–confirmation model (ECM), we tested a structural equation model using survey data from 512 respondents. The findings indicate that algorithmic mediation enhances both expectation confirmation and virtual engagement. Virtual engagement significantly fosters trust and directly predicts travel intentions, while trust emerges as the strongest mediator linking digital interactions to behavioral outcomes. Multi-group analysis reveals that prospective travelers rely more on virtual engagement to build trust than experienced travelers, highlighting the conditional role of prior travel experience. This research contributes to the smart tourism literature by integrating algorithmic mediation and virtual engagement into the ECM framework, emphasizing their crucial roles in shaping Generation Z tourists' digital decision-making. Practical implications suggest that tourism platforms should refine algorithmic design and interactive features to strengthen trust and effectively convert online engagement into actual travel behaviors.
- 1 September 2025
- Published by Learning Gate in Edelweiss Applied Science and Technology
- Vol. 9 (9) , 486-503
Open AccessAlgorithmic mediation and virtual engagement in smart tourism: An expectation–confirmation model of trust and travel intention among Gen Z tourists in Thailand
Tourism in the digital era is increasingly influenced by algorithmic recommendation systems and immersive online interactions. This study examines how algorithmic mediation and virtual engagement jointly affect Chinese Generation Z tourists' trust and travel intentions in Thailand, a leading destination for smart tourism. Guided by the expectation–confirmation model (ECM), we tested a structural equation model using survey data from 512 respondents. The findings indicate that algorithmic mediation enhances both expectation confirmation and virtual engagement. Virtual engagement significantly fosters trust and directly predicts travel intentions, while trust emerges as the strongest mediator linking digital interactions to behavioral outcomes. Multi-group analysis reveals that prospective travelers rely more on virtual engagement to build trust than experienced travelers, highlighting the conditional role of prior travel experience. This research contributes to the smart tourism literature by integrating algorithmic mediation and virtual engagement into the ECM framework, emphasizing their crucial roles in shaping Generation Z tourists' digital decision-making. Practical implications suggest that tourism platforms should refine algorithmic design and interactive features to strengthen trust and effectively convert online engagement into actual travel behaviors.
- Open Access
- 19 August 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (16) , 2665
Int.2D-3D-CNN: Integrated 2D and 3D Convolutional Neural Networks for Video Violence Recognition
Intelligent video analysis tools have advanced significantly, with numerous cameras installed in various locations to enhance security and monitor unusual events. However, the effective detection and monitoring of violent incidents often depend on manual effort and time-consuming analysis of recorded footage, which can delay timely interventions. Deep learning has emerged as a powerful approach for extracting critical features essential to identifying and classifying violent behavior, enabling the development of accurate and scalable models across diverse domains. This study presents the Int.2D-3D-CNN architecture, which integrates a two-dimensional convolutional neural network (2D-CNN) and 3D-CNNs for video-based violence recognition. Compared to traditional 2D-CNN and 3D-CNN models, the proposed Int.2D-3D-CNN model presents improved performance on the Hockey Fight, Movie, and Violent Flows datasets. The architecture captures both static and dynamic characteristics of violent scenes by integrating spatial and temporal information. Specifically, the 2D-CNN component employs lightweight MobileNetV1 and MobileNetV2 to extract spatial features from individual frames, while a simplified 3D-CNN module with a single 3D convolution layer captures motion and temporal dependencies across sequences. Evaluation results highlight the robustness of the proposed model in accurately distinguishing violent from non-violent videos under diverse conditions. The Int.2D-3D-CNN model achieved accuracies of 98%, 100%, and 98% on the Hockey Fight, Movie, and Violent Flows datasets, respectively, indicating strong potential for violence recognition applications.
- 19 August 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (16) , 2665
Open AccessInt.2D-3D-CNN: Integrated 2D and 3D Convolutional Neural Networks for Video Violence Recognition
Intelligent video analysis tools have advanced significantly, with numerous cameras installed in various locations to enhance security and monitor unusual events. However, the effective detection and monitoring of violent incidents often depend on manual effort and time-consuming analysis of recorded footage, which can delay timely interventions. Deep learning has emerged as a powerful approach for extracting critical features essential to identifying and classifying violent behavior, enabling the development of accurate and scalable models across diverse domains. This study presents the Int.2D-3D-CNN architecture, which integrates a two-dimensional convolutional neural network (2D-CNN) and 3D-CNNs for video-based violence recognition. Compared to traditional 2D-CNN and 3D-CNN models, the proposed Int.2D-3D-CNN model presents improved performance on the Hockey Fight, Movie, and Violent Flows datasets. The architecture captures both static and dynamic characteristics of violent scenes by integrating spatial and temporal information. Specifically, the 2D-CNN component employs lightweight MobileNetV1 and MobileNetV2 to extract spatial features from individual frames, while a simplified 3D-CNN module with a single 3D convolution layer captures motion and temporal dependencies across sequences. Evaluation results highlight the robustness of the proposed model in accurately distinguishing violent from non-violent videos under diverse conditions. The Int.2D-3D-CNN model achieved accuracies of 98%, 100%, and 98% on the Hockey Fight, Movie, and Violent Flows datasets, respectively, indicating strong potential for violence recognition applications.
- Open Access
Multi-language Video Subtitle Dataset for Image-based Text Recognition
The Multi-language Video Subtitle Dataset is a comprehensive collection designed to support research in text recognition across multiple languages. This dataset includes 4,224 subtitle images extracted from 24 videos sourced from online platforms. It features a wide variety of characters, including Thai consonants, vowels, tone marks, punctuation marks, numerals, Roman characters, and Arabic numerals. With 157 unique characters, the dataset provides a resource for addressing challenges in text recognition within complex backgrounds. It addresses the growing need for high-quality, multilingual text recognition data, particularly as videos with embedded subtitles become increasingly dominant on platforms like YouTube and Facebook. The variability in text length, font, and placement within these images adds complexity, offering a valuable resource for developing and evaluating deep learning models. The dataset facilitates accurate text transcription from video content while providing a foundation for improving computational efficiency in text recognition systems. As a result, it holds significant potential to drive advancements in research and innovation across various computer science disciplines, including artificial intelligence, deep learning, computer vision, and pattern recognition.
Open AccessMulti-language Video Subtitle Dataset for Image-based Text Recognition
The Multi-language Video Subtitle Dataset is a comprehensive collection designed to support research in text recognition across multiple languages. This dataset includes 4,224 subtitle images extracted from 24 videos sourced from online platforms. It features a wide variety of characters, including Thai consonants, vowels, tone marks, punctuation marks, numerals, Roman characters, and Arabic numerals. With 157 unique characters, the dataset provides a resource for addressing challenges in text recognition within complex backgrounds. It addresses the growing need for high-quality, multilingual text recognition data, particularly as videos with embedded subtitles become increasingly dominant on platforms like YouTube and Facebook. The variability in text length, font, and placement within these images adds complexity, offering a valuable resource for developing and evaluating deep learning models. The dataset facilitates accurate text transcription from video content while providing a foundation for improving computational efficiency in text recognition systems. As a result, it holds significant potential to drive advancements in research and innovation across various computer science disciplines, including artificial intelligence, deep learning, computer vision, and pattern recognition.
- Open Access
AIWR: Aerial Image Water Resource Dataset for Segmentation Analysis
Effective water resource management is crucial in agricultural regions like northeastern Thailand, where limited water retention in sandy soils poses significant challenges. In response to this issue, the Aerial Image Water Resource (AIWR) dataset was developed, comprising 800 aerial images focused on natural and artificial water bodies in this region. The dataset was created using Bing Maps and follows the standards of the Fundamental Geographic Data Set (FGDS). It includes ground truth annotations validated by experts in remote sensing, making it an invaluable resource for researchers in geoinformatics, computer vision, and artificial intelligence. The AIWR dataset presents considerable challenges, such as segmentation due to variations in the size, color, shape, and similarity of water bodies, which often resemble other land use categories. The objective of the proposed dataset is to explore advanced AI-driven methods for water body segmentation, addressing the unique challenges posed by the dataset complexity and limited size. This dataset and related research contribute to the development of novel algorithms for water management, supporting sustainable agricultural practices in regions facing similar challenges.
Open AccessAIWR: Aerial Image Water Resource Dataset for Segmentation Analysis
Effective water resource management is crucial in agricultural regions like northeastern Thailand, where limited water retention in sandy soils poses significant challenges. In response to this issue, the Aerial Image Water Resource (AIWR) dataset was developed, comprising 800 aerial images focused on natural and artificial water bodies in this region. The dataset was created using Bing Maps and follows the standards of the Fundamental Geographic Data Set (FGDS). It includes ground truth annotations validated by experts in remote sensing, making it an invaluable resource for researchers in geoinformatics, computer vision, and artificial intelligence. The AIWR dataset presents considerable challenges, such as segmentation due to variations in the size, color, shape, and similarity of water bodies, which often resemble other land use categories. The objective of the proposed dataset is to explore advanced AI-driven methods for water body segmentation, addressing the unique challenges posed by the dataset complexity and limited size. This dataset and related research contribute to the development of novel algorithms for water management, supporting sustainable agricultural practices in regions facing similar challenges.
- Open Access
EcoCropsAID: Economic Crops Aerial Image Dataset for Land Use Classification
The EcoCropsAID dataset is a comprehensive collection of 5,400 aerial images captured between 2014 and 2018 using the Google Earth application. This dataset focuses on five key economic crops in Thailand: rice, sugarcane, cassava, rubber, and longan. The images were collected at various crop growth stages: early cultivation, growth, and harvest, resulting in significant variability within each category and similarities across different categories. These variations, coupled with differences in resolution, color, and contrast introduced by multiple remote imaging sensors, present substantial challenges for land use classification. The dataset is an interdisciplinary resource that spans multiple research domains, including remote sensing, geoinformatics, artificial intelligence, and computer vision. The unique features of the EcoCropsAID dataset offer opportunities for researchers to explore novel approaches, such as extracting spatial and temporal features, developing deep learning architectures, and implementing transformer-based models. The EcoCropsAID dataset provides a valuable platform for advancing research in land use classification, with implications for optimizing agricultural practices and enhancing sustainable development. This study explicitly investigates the use of deep learning algorithms to classify economic crop areas in northeastern Thailand, utilizing satellite imagery to address the challenges posed by diverse patterns and similarities across categories.
Open AccessEcoCropsAID: Economic Crops Aerial Image Dataset for Land Use Classification
The EcoCropsAID dataset is a comprehensive collection of 5,400 aerial images captured between 2014 and 2018 using the Google Earth application. This dataset focuses on five key economic crops in Thailand: rice, sugarcane, cassava, rubber, and longan. The images were collected at various crop growth stages: early cultivation, growth, and harvest, resulting in significant variability within each category and similarities across different categories. These variations, coupled with differences in resolution, color, and contrast introduced by multiple remote imaging sensors, present substantial challenges for land use classification. The dataset is an interdisciplinary resource that spans multiple research domains, including remote sensing, geoinformatics, artificial intelligence, and computer vision. The unique features of the EcoCropsAID dataset offer opportunities for researchers to explore novel approaches, such as extracting spatial and temporal features, developing deep learning architectures, and implementing transformer-based models. The EcoCropsAID dataset provides a valuable platform for advancing research in land use classification, with implications for optimizing agricultural practices and enhancing sustainable development. This study explicitly investigates the use of deep learning algorithms to classify economic crop areas in northeastern Thailand, utilizing satellite imagery to address the challenges posed by diverse patterns and similarities across categories.
- Open Access
- 11 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 132369-132389
Robust Model Selection for Plant Leaf Image Recognition Based on Evolutionary Ant Colony Optimization With Learning Rate Schedule
Selecting optimal deep learning models is often a time-consuming process. To address this challenge, we propose a novel variant of the ant colony optimization (ACO) algorithm. This approach is designed to enhance model selection across various deep learning architectures, with a particular focus on leaf classification tasks. We introduce a new ACO technique specifically tailored for selecting robust models within convolutional neural networks (CNNs). These models are then integrated into an ensemble learning framework known as ensemble CNNs. A distinguishing feature of our proposed evolutionary ACO algorithm is its ability to consistently identify a set of robust CNN models in each iteration. This capability is facilitated by an innovative fitness function and an adaptive learning rate schedule embedded within the ACO algorithm, which optimizes pheromone distribution. Unlike the original ACO algorithm, which consistently selects the same CNN model, our evolutionary approach enables the dynamic discovery of new CNN models. To validate our method, we conducted experiments on two plant leaf datasets: Mulberry and Turkey-plant. Our comparison with existing methods, specifically the ant colony system (ACS) and the max-min ant system (MMAS), demonstrated that the MMAS algorithm outperformed the ACS algorithm. Furthermore, we explored three ensemble learning techniques: unweighted average, weighted average, and cost-sensitive learning. The weighted average method emerged as the most effective ensemble approach, with its parameters determined through a grid search process. The results indicate that the evolutionary ACO algorithm not only facilitates the selection of robust deep learning models but also achieves superior performance compared to the original ACO algorithm when applied to the Mulberry leaf and Turkey-plant datasets.
- 11 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 132369-132389
Open AccessRobust Model Selection for Plant Leaf Image Recognition Based on Evolutionary Ant Colony Optimization With Learning Rate Schedule
Selecting optimal deep learning models is often a time-consuming process. To address this challenge, we propose a novel variant of the ant colony optimization (ACO) algorithm. This approach is designed to enhance model selection across various deep learning architectures, with a particular focus on leaf classification tasks. We introduce a new ACO technique specifically tailored for selecting robust models within convolutional neural networks (CNNs). These models are then integrated into an ensemble learning framework known as ensemble CNNs. A distinguishing feature of our proposed evolutionary ACO algorithm is its ability to consistently identify a set of robust CNN models in each iteration. This capability is facilitated by an innovative fitness function and an adaptive learning rate schedule embedded within the ACO algorithm, which optimizes pheromone distribution. Unlike the original ACO algorithm, which consistently selects the same CNN model, our evolutionary approach enables the dynamic discovery of new CNN models. To validate our method, we conducted experiments on two plant leaf datasets: Mulberry and Turkey-plant. Our comparison with existing methods, specifically the ant colony system (ACS) and the max-min ant system (MMAS), demonstrated that the MMAS algorithm outperformed the ACS algorithm. Furthermore, we explored three ensemble learning techniques: unweighted average, weighted average, and cost-sensitive learning. The weighted average method emerged as the most effective ensemble approach, with its parameters determined through a grid search process. The results indicate that the evolutionary ACO algorithm not only facilitates the selection of robust deep learning models but also achieves superior performance compared to the original ACO algorithm when applied to the Mulberry leaf and Turkey-plant datasets.
- Open Access
- 9 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 52237-52248
Improving Neural Network-Based Multi-Label Classification With Pattern Loss Penalties
This research work introduces two novel loss functions, pattern-loss (POL) and label similarity-based instance modeling (LSIM), for improving the performance of multi-label classification using artificial neural network-based techniques. These loss functions incorporate additional optimization constraints based on the distribution of multi-label class patterns and the similarity of data instances. By integrating these patterns during the network training process, the trained model is tuned to align with the existing patterns in the training data. The proposed approach decomposes the loss function into two components: the cross entropy loss and the pattern loss derived from the distribution of class-label patterns. Experimental evaluations were conducted on eight standard datasets, comparing the proposed methods with three existing techniques.The results demonstrate the effectiveness of the proposed approach, with POL and LSIM consistently achieving superior accuracy performance compared to the benchmark methods.
- 9 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 52237-52248
Open AccessImproving Neural Network-Based Multi-Label Classification With Pattern Loss Penalties
This research work introduces two novel loss functions, pattern-loss (POL) and label similarity-based instance modeling (LSIM), for improving the performance of multi-label classification using artificial neural network-based techniques. These loss functions incorporate additional optimization constraints based on the distribution of multi-label class patterns and the similarity of data instances. By integrating these patterns during the network training process, the trained model is tuned to align with the existing patterns in the training data. The proposed approach decomposes the loss function into two components: the cross entropy loss and the pattern loss derived from the distribution of class-label patterns. Experimental evaluations were conducted on eight standard datasets, comparing the proposed methods with three existing techniques.The results demonstrate the effectiveness of the proposed approach, with POL and LSIM consistently achieving superior accuracy performance compared to the benchmark methods.