Pub Date : 2025-03-06DOI: 10.1109/ACCESS.2025.3549101
Mu Ren;Shiliang Wang;Bohan Liang;Cui Wei
In this paper, a multi-origin and multi-destination maximum flow model is proposed by integrating the objective programming method with graph theory. Through Model 1, low-cost design of drainage pipeline networks can be achieved. For instance, in the experiments, certain pipelines such as Pipeline 9, Pipeline 20, and Pipeline 27 can be excluded during the initial construction phase while still maintaining effective drainage performance and reducing costs. Model 2 enables the monitoring of drainage capacity at various nodes within the existing drainage system. For example, under a specific rainfall intensity, it was observed that certain nodes, such as Nodes 9~16, Node 45, and Node 48, exhibit poor drainage capacity, necessitating precautionary measures. Model 3 focuses on the expansion planning of the drainage system, ensuring its normal operation under certain rainfall intensities. The results based on a specific area in Changchun City demonstrate that the proposed models not only possess strong practical application value in the design of urban drainage systems but also have broader applicability in other fields such as traffic flow and logistics supply chain planning. This study opens up new avenues for model construction and enriches the practical application effects of mixed-integer programming models.
{"title":"Maximum Flow Model With Multiple Origin and Destination and Its Application in Designing Urban Drainage Systems","authors":"Mu Ren;Shiliang Wang;Bohan Liang;Cui Wei","doi":"10.1109/ACCESS.2025.3549101","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3549101","url":null,"abstract":"In this paper, a multi-origin and multi-destination maximum flow model is proposed by integrating the objective programming method with graph theory. Through Model 1, low-cost design of drainage pipeline networks can be achieved. For instance, in the experiments, certain pipelines such as Pipeline 9, Pipeline 20, and Pipeline 27 can be excluded during the initial construction phase while still maintaining effective drainage performance and reducing costs. Model 2 enables the monitoring of drainage capacity at various nodes within the existing drainage system. For example, under a specific rainfall intensity, it was observed that certain nodes, such as Nodes 9~16, Node 45, and Node 48, exhibit poor drainage capacity, necessitating precautionary measures. Model 3 focuses on the expansion planning of the drainage system, ensuring its normal operation under certain rainfall intensities. The results based on a specific area in Changchun City demonstrate that the proposed models not only possess strong practical application value in the design of urban drainage systems but also have broader applicability in other fields such as traffic flow and logistics supply chain planning. This study opens up new avenues for model construction and enriches the practical application effects of mixed-integer programming models.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"44403-44417"},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10916634","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143619077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-06DOI: 10.1109/ACCESS.2025.3548631
Felipe Ramos;Alexandre Costa;Mirko Perkusich;Luiz Silva;Dalton Valadares;Ademar de Sousa Neto;Felipe Cunha;Hyggo Almeida;Angelo Perkusich
Context: Agile software development, particularly Scrum, enables teams to manage evolving requirements by emphasizing face-to-face communication and incremental deliveries. Although effective in addressing functional requirements, agile methods often overlook non-functional requirements during the initial stages of software projects, potentially leading to cost overruns on software and hardware and project failures exceeding 60%. Objective: In this article, we introduce a data-driven recommendation system to assist Scrum teams in eliciting NFRs effectively and early in the development lifecycle. Method: Our proposed solution applies the k-nearest neighbors algorithm to recommend non-functional requirements by leveraging historical project data structured through a taxonomy of user stories. We evaluated the system through offline experiments under the cross-validation protocol, utilizing datasets from 13 real-world projects. Results: Our recommendation system achieved an F-measure of up to 79%, demonstrating its ability to provide accurate and context-aware non-functional requirements suggestions. Conclusion: These findings suggest that our solution supports agile teams by automating non-functional requirement elicitation and enhancing decision-making processes, thereby addressing critical gaps in non-functional requirement integration within Scrum-based projects.
{"title":"A Data-Driven Recommendation System for Enhancing Non-Functional Requirements Elicitation in Scrum-Based Projects","authors":"Felipe Ramos;Alexandre Costa;Mirko Perkusich;Luiz Silva;Dalton Valadares;Ademar de Sousa Neto;Felipe Cunha;Hyggo Almeida;Angelo Perkusich","doi":"10.1109/ACCESS.2025.3548631","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548631","url":null,"abstract":"<bold>Context:</b> Agile software development, particularly Scrum, enables teams to manage evolving requirements by emphasizing face-to-face communication and incremental deliveries. Although effective in addressing functional requirements, agile methods often overlook non-functional requirements during the initial stages of software projects, potentially leading to cost overruns on software and hardware and project failures exceeding 60%. <bold>Objective:</b> In this article, we introduce a data-driven recommendation system to assist Scrum teams in eliciting NFRs effectively and early in the development lifecycle. <bold>Method:</b> Our proposed solution applies the k-nearest neighbors algorithm to recommend non-functional requirements by leveraging historical project data structured through a taxonomy of user stories. We evaluated the system through offline experiments under the cross-validation protocol, utilizing datasets from 13 real-world projects. <bold>Results:</b> Our recommendation system achieved an F-measure of up to 79%, demonstrating its ability to provide accurate and context-aware non-functional requirements suggestions. <bold>Conclusion:</b> These findings suggest that our solution supports agile teams by automating non-functional requirement elicitation and enhancing decision-making processes, thereby addressing critical gaps in non-functional requirement integration within Scrum-based projects.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"44000-44023"},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10915619","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621858","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-06DOI: 10.1109/ACCESS.2025.3549312
Jae H. Park;Vijay K. Madisetti
While the swift advancement of cloud-based Large Language Models (LLMs) has significantly increased the efficiency and automation in business processes, it has also introduced considerable privacy concerns regarding Personally Identifiable Information (PII) and other protected data in multimodal forms, such as text, video, or images, being exported, potentially insecurely, outside the corporate environments. Although traditional anonymization-based techniques can alleviate these risks in offline applications, such as summarization or classification, incorporating it into online LLM workflows poses substantial challenges, particularly when these workflows encompass real-time transactions involving multiple stakeholders, as commonly observed in multi-agent generative AI applications. This study explores these challenges and proposes novel context-aware privacy frameworks and methods to address these issues. We employ a local privacy-focused gatekeeper LLM to contextually pseudonymize PII and assign unique identifiers as part of a new mapping process, thereby facilitating re-identification in real-time operations while safeguarding privacy when interacting with cloud-based LLMs. Our proposed methodologies and frameworks adeptly integrate privacy considerations into LLM and LLM Agent workflows, preserving both privacy and data utility while maintaining operational efficiency and utility comparable to non-anonymized generative AI processes.
{"title":"CAPRI: A Context-Aware Privacy Framework for Multi-Agent Generative AI Applications","authors":"Jae H. Park;Vijay K. Madisetti","doi":"10.1109/ACCESS.2025.3549312","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3549312","url":null,"abstract":"While the swift advancement of cloud-based Large Language Models (LLMs) has significantly increased the efficiency and automation in business processes, it has also introduced considerable privacy concerns regarding Personally Identifiable Information (PII) and other protected data in multimodal forms, such as text, video, or images, being exported, potentially insecurely, outside the corporate environments. Although traditional anonymization-based techniques can alleviate these risks in offline applications, such as summarization or classification, incorporating it into online LLM workflows poses substantial challenges, particularly when these workflows encompass real-time transactions involving multiple stakeholders, as commonly observed in multi-agent generative AI applications. This study explores these challenges and proposes novel context-aware privacy frameworks and methods to address these issues. We employ a local privacy-focused gatekeeper LLM to contextually pseudonymize PII and assign unique identifiers as part of a new mapping process, thereby facilitating re-identification in real-time operations while safeguarding privacy when interacting with cloud-based LLMs. Our proposed methodologies and frameworks adeptly integrate privacy considerations into LLM and LLM Agent workflows, preserving both privacy and data utility while maintaining operational efficiency and utility comparable to non-anonymized generative AI processes.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"43168-43177"},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10916629","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143632258","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-06DOI: 10.1109/ACCESS.2025.3548572
Hangwook Jeong;Minseon Park;Min-Woo Kwon
In this study, we propose a feedback field effect transistor (FBFET) in-memory computing device based on the string of NAND flash array. By combining the polysilicon-blocking oxide-silicon nitride-tunneling oxide-silicon (SONOS) cell structure and N+-P$^{-}$ -N$^{-}$ -P+ body doping structure of FBFET, the device is designed in a string structure. Our in-memory computing FBFET can perform Boolean logical operations by utilizing the threshold voltage modulation characteristics of the FBFET by adjusting the control gate voltage. Additionally, we present optimized hot carrier injection conditions tailored for the in-memory computing FBFET, enabling the successful writing of results for Boolean logic operations: OR, AND, NOR, NAND. These findings contribute a novel mechanism and direction for the field of in-memory computing research.
{"title":"Development of In-Memory Computing Device Using Positive Feedback Field Effect Transistor Based on NAND Flash Array","authors":"Hangwook Jeong;Minseon Park;Min-Woo Kwon","doi":"10.1109/ACCESS.2025.3548572","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548572","url":null,"abstract":"In this study, we propose a feedback field effect transistor (FBFET) in-memory computing device based on the string of NAND flash array. By combining the polysilicon-blocking oxide-silicon nitride-tunneling oxide-silicon (SONOS) cell structure and N+-P<inline-formula> <tex-math>$^{-}$ </tex-math></inline-formula>-N<inline-formula> <tex-math>$^{-}$ </tex-math></inline-formula>-P+ body doping structure of FBFET, the device is designed in a string structure. Our in-memory computing FBFET can perform Boolean logical operations by utilizing the threshold voltage modulation characteristics of the FBFET by adjusting the control gate voltage. Additionally, we present optimized hot carrier injection conditions tailored for the in-memory computing FBFET, enabling the successful writing of results for Boolean logic operations: OR, AND, NOR, NAND. These findings contribute a novel mechanism and direction for the field of in-memory computing research.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"45449-45457"},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10915680","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143645276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1109/ACCESS.2025.3548134
Andrea Eirale;Mauro Martini;Marcello Chiaberge
Collaborative and companion robots are at the forefront of technological innovation, transforming human-robot interaction to address a wide range of tasks and activities. This review provides a comprehensive examination of the current state of research on human following and guidance with autonomous mobile robots. Covering the evolution of research from the inception to the latest advancements in this area, we categorize existing literature based on various attributes including fields of application, technologies employed, and social acceptability. We critically analyze and compare state-of-the-art approaches in perception, tracking, planning, control, and human-robot interaction, highlighting their effectiveness and feasibility. We further classify studies based on application domains where person following and guiding tasks are particularly impactful, such as healthcare, personal assistance, logistics, and tour guiding. We identify persistent challenges and outline open problems, offering recommendations for future research directions. Our review aims to serve as a foundational reference for researchers and practitioners, fostering continued innovation and development in the deployment of autonomous robots for human following and guidance.
{"title":"Human Following and Guidance by Autonomous Mobile Robots: A Comprehensive Review","authors":"Andrea Eirale;Mauro Martini;Marcello Chiaberge","doi":"10.1109/ACCESS.2025.3548134","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548134","url":null,"abstract":"Collaborative and companion robots are at the forefront of technological innovation, transforming human-robot interaction to address a wide range of tasks and activities. This review provides a comprehensive examination of the current state of research on human following and guidance with autonomous mobile robots. Covering the evolution of research from the inception to the latest advancements in this area, we categorize existing literature based on various attributes including fields of application, technologies employed, and social acceptability. We critically analyze and compare state-of-the-art approaches in perception, tracking, planning, control, and human-robot interaction, highlighting their effectiveness and feasibility. We further classify studies based on application domains where person following and guiding tasks are particularly impactful, such as healthcare, personal assistance, logistics, and tour guiding. We identify persistent challenges and outline open problems, offering recommendations for future research directions. Our review aims to serve as a foundational reference for researchers and practitioners, fostering continued innovation and development in the deployment of autonomous robots for human following and guidance.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"42214-42253"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10910101","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143594392","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Multi-machine collaboration in agricultural machinery is a key focus in current research, with task allocation being an indispensable component. However, the current optimization objectives for task allocation in agricultural machinery are mostly confined to travel distance or time, aiming to balance task distribution. These methods are not suitable for emerging electric agricultural machinery, especially when operating in hilly areas. To address these limitations, this study proposed a task allocation method optimized for energy consumption, specifically for weeding robots in hilly orchards. Initially, drones were employed to obtain the Digital Surface Model (DSM) and orthophotos of the orchard test area. After processing the data through vegetation filtering, DEM construction, and slope analysis, slope information of the surface was derived. An electronic map of the orchard reflecting this slope information was then generated. Subsequently, the task allocation problem for weeding robots in hilly orchards was defined. A mathematical model was then established with energy consumption as the optimization objective. Finally, a Golden Kepler Optimization Algorithm (GKOA) was developed and tested through simulations using real data from the test area. The results indicated that, compared to Particle Swarm Optimization (PSO), Sparrow Search Algorithm (SSA), Whale Optimization Algorithm (WOA), and Kepler Optimization Algorithm (KOA), GKOA reduced the optimal solution cost by 10.3%, 8.2%, 7.0%, and 4.5%, respectively. This task allocation method was able to achieve the optimal task allocation plan with lower travel energy consumption costs and a higher balance in task distribution, whether for all plots in the orchard or nested plots.
{"title":"Development of an Improved KOA Algorithm for Solving Task Allocation in Hilly Orchards With Weeding Robots","authors":"Xiaolin Xie;Hang Jin;Heng Wang;Man Xu;Cheng Zhang;Xin Jin;Zhihong Zhang","doi":"10.1109/ACCESS.2025.3548162","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548162","url":null,"abstract":"Multi-machine collaboration in agricultural machinery is a key focus in current research, with task allocation being an indispensable component. However, the current optimization objectives for task allocation in agricultural machinery are mostly confined to travel distance or time, aiming to balance task distribution. These methods are not suitable for emerging electric agricultural machinery, especially when operating in hilly areas. To address these limitations, this study proposed a task allocation method optimized for energy consumption, specifically for weeding robots in hilly orchards. Initially, drones were employed to obtain the Digital Surface Model (DSM) and orthophotos of the orchard test area. After processing the data through vegetation filtering, DEM construction, and slope analysis, slope information of the surface was derived. An electronic map of the orchard reflecting this slope information was then generated. Subsequently, the task allocation problem for weeding robots in hilly orchards was defined. A mathematical model was then established with energy consumption as the optimization objective. Finally, a Golden Kepler Optimization Algorithm (GKOA) was developed and tested through simulations using real data from the test area. The results indicated that, compared to Particle Swarm Optimization (PSO), Sparrow Search Algorithm (SSA), Whale Optimization Algorithm (WOA), and Kepler Optimization Algorithm (KOA), GKOA reduced the optimal solution cost by 10.3%, 8.2%, 7.0%, and 4.5%, respectively. This task allocation method was able to achieve the optimal task allocation plan with lower travel energy consumption costs and a higher balance in task distribution, whether for all plots in the orchard or nested plots.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"44184-44195"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10910132","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621552","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1109/ACCESS.2025.3548170
Wenhao Zeng;Gongbing Su;Zixuan Su;Rui Li;Jun Chen
Point-cloud registration and stitching are important topics in the field of robot navigation and 3D reconstruction, e.g., the accuracy of point cloud registration and stitching in robot navigation directly affects the accuracy of map construction. Many researchers have proposed various algorithms for deep learning-based point cloud registration and stitching methods with good performance, and although there are end-to-end methods that have made progress, they still have limitations in local feature fusion efficiency and geometric detail retention. To address this issue, a fusion algorithm for registration and stitching based on a GCN-PRFNet point cloud is proposed. The network has a feature extraction module, a point cloud registration module, and a point cloud splicing and fusion module. GCN-PRFNet can efficiently handle the task of point cloud registration and splicing and fusion in partially overlapping regions and is robust to noise. The model is trained on the ModelNet40 dataset, and its registration and splicing accuracies are improved by 53.9%, 20.1%, 8.3%, 12.2%, 6.1% and 1.8% when compared with the traditional iterative closest point and learning-based PointNetLK, DGCNN, RPM-Net, DCP, and PointViG methods. This indicates that the constructed model is effective in point cloud registration and splicing. Meanwhile, point-cloud registration and splicing tests were performed on five self-constructed artefact datasets, and their registration and splicing accuracies were over 90%, indicating that the constructed end-to-end point-cloud registration and splicing model is considerably effective in real-world application scenarios.
{"title":"Research on Point Cloud Registration and Stitching Fusion Algorithm Based on GCN-PRFNet","authors":"Wenhao Zeng;Gongbing Su;Zixuan Su;Rui Li;Jun Chen","doi":"10.1109/ACCESS.2025.3548170","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548170","url":null,"abstract":"Point-cloud registration and stitching are important topics in the field of robot navigation and 3D reconstruction, e.g., the accuracy of point cloud registration and stitching in robot navigation directly affects the accuracy of map construction. Many researchers have proposed various algorithms for deep learning-based point cloud registration and stitching methods with good performance, and although there are end-to-end methods that have made progress, they still have limitations in local feature fusion efficiency and geometric detail retention. To address this issue, a fusion algorithm for registration and stitching based on a GCN-PRFNet point cloud is proposed. The network has a feature extraction module, a point cloud registration module, and a point cloud splicing and fusion module. GCN-PRFNet can efficiently handle the task of point cloud registration and splicing and fusion in partially overlapping regions and is robust to noise. The model is trained on the ModelNet40 dataset, and its registration and splicing accuracies are improved by 53.9%, 20.1%, 8.3%, 12.2%, 6.1% and 1.8% when compared with the traditional iterative closest point and learning-based PointNetLK, DGCNN, RPM-Net, DCP, and PointViG methods. This indicates that the constructed model is effective in point cloud registration and splicing. Meanwhile, point-cloud registration and splicing tests were performed on five self-constructed artefact datasets, and their registration and splicing accuracies were over 90%, indicating that the constructed end-to-end point-cloud registration and splicing model is considerably effective in real-world application scenarios.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"43384-43397"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10910120","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1109/ACCESS.2025.3548563
Dale Black;Wenbo Li;Qiyu Zhang;Sabee Molloi
Euclidean distance transforms are fundamental in image processing and computer vision, with critical applications in medical image analysis and computer graphics. However, existing implementations often lack performance, flexibility, or cross-platform compatibility. This paper introduces a novel approach to accelerating Euclidean distance transforms using hardware-agnostic GPU acceleration, multi-threading, and cross-language support. Our method, implemented in Julia with Python bindings, supports multiple GPU platforms including NVIDIA CUDA, AMD ROCm, Apple Metal, and Intel oneAPI. Benchmarks demonstrate substantial performance improvements, achieving speedups by a factor of 250 for 2D and a factor of 400 for 3D transforms compared to optimized CPU implementations. We showcase the impact of our approach through two real-world applications: accelerating the Hausdorff distance loss function for medical image segmentation, achieving a 7.4-fold improvement in processing speed with enhanced accuracy, and enhancing a GPU-optimized distance transform-based skeletonization algorithm with performance gains up to a factor of 88. Our open-source implementation provides a flexible, high-performance solution for exact Euclidean distance transforms, advancing the state-of-the-art in medical image analysis and computer vision.
{"title":"Accelerating Euclidean Distance Transforms: A Fast and Flexible Approach With Multi-Vendor GPU, Multi-Threading, and Multi-Language Support","authors":"Dale Black;Wenbo Li;Qiyu Zhang;Sabee Molloi","doi":"10.1109/ACCESS.2025.3548563","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548563","url":null,"abstract":"Euclidean distance transforms are fundamental in image processing and computer vision, with critical applications in medical image analysis and computer graphics. However, existing implementations often lack performance, flexibility, or cross-platform compatibility. This paper introduces a novel approach to accelerating Euclidean distance transforms using hardware-agnostic GPU acceleration, multi-threading, and cross-language support. Our method, implemented in Julia with Python bindings, supports multiple GPU platforms including NVIDIA CUDA, AMD ROCm, Apple Metal, and Intel oneAPI. Benchmarks demonstrate substantial performance improvements, achieving speedups by a factor of 250 for 2D and a factor of 400 for 3D transforms compared to optimized CPU implementations. We showcase the impact of our approach through two real-world applications: accelerating the Hausdorff distance loss function for medical image segmentation, achieving a 7.4-fold improvement in processing speed with enhanced accuracy, and enhancing a GPU-optimized distance transform-based skeletonization algorithm with performance gains up to a factor of 88. Our open-source implementation provides a flexible, high-performance solution for exact Euclidean distance transforms, advancing the state-of-the-art in medical image analysis and computer vision.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"44636-44649"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10912438","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621659","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1109/ACCESS.2025.3548510
Jorge K. S. Kamassury;Henrique Pickler;Filipe R. Cordeiro;Danilo Silva
The detrimental impact of noisy labels on the generalization performance of deep neural networks has sparked research interest in learning with noisy labels (LNL). Among the various methods proposed to mitigate this effect, the Co-Teaching method, characterized by co-training with the small-loss criterion, is one of the most established approaches and is widely employed as a key component in recent LNL methods. Although Co-Teaching can mitigate the overfitting effect, it still remains, especially in scenarios with high rates of label noise in datasets. Strategies from the LNL literature to address this typically include the use of disagreement techniques and alternative loss functions. In this paper, we propose the Cyclic Co-Teaching (CCT) method, which employs cyclic variations in the learning rate and sample retention rate at the mini-batch level, along with a checkpoint mechanism that ensures that training in subsequent cycles always resumes from the best models obtained so far. For optimizing the method, we developed a framework that incorporates a pre-training phase to obtain an optimized vanilla model used to initialize CCT model weights, and a transparent univariate optimization strategy for hyperparameters that does not necessarily require a clean validation set. Experimental results on synthetic and real-world datasets, under different types and levels of noise and employing various neural network architectures, demonstrate that CCT outperforms several state-of-the-art LNL methods in most evaluated scenarios.
{"title":"CCT: A Cyclic Co-Teaching Approach to Train Deep Neural Networks With Noisy Labels","authors":"Jorge K. S. Kamassury;Henrique Pickler;Filipe R. Cordeiro;Danilo Silva","doi":"10.1109/ACCESS.2025.3548510","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548510","url":null,"abstract":"The detrimental impact of noisy labels on the generalization performance of deep neural networks has sparked research interest in learning with noisy labels (LNL). Among the various methods proposed to mitigate this effect, the Co-Teaching method, characterized by co-training with the small-loss criterion, is one of the most established approaches and is widely employed as a key component in recent LNL methods. Although Co-Teaching can mitigate the overfitting effect, it still remains, especially in scenarios with high rates of label noise in datasets. Strategies from the LNL literature to address this typically include the use of disagreement techniques and alternative loss functions. In this paper, we propose the Cyclic Co-Teaching (CCT) method, which employs cyclic variations in the learning rate and sample retention rate at the mini-batch level, along with a checkpoint mechanism that ensures that training in subsequent cycles always resumes from the best models obtained so far. For optimizing the method, we developed a framework that incorporates a pre-training phase to obtain an optimized vanilla model used to initialize CCT model weights, and a transparent univariate optimization strategy for hyperparameters that does not necessarily require a clean validation set. Experimental results on synthetic and real-world datasets, under different types and levels of noise and employing various neural network architectures, demonstrate that CCT outperforms several state-of-the-art LNL methods in most evaluated scenarios.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"43843-43860"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10912480","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1109/ACCESS.2025.3548273
Di Ding;Tianliang Yao;Haoyu Wang;Xusen Sun;Rong Luo
Transcatheter mitral valve procedures have transformed the treatment of mitral regurgitation and stenosis by providing less invasive alternatives to conventional open-heart surgery. However, they introduce stringent requirements for catheter navigation, force modulation, and real-time imaging in a highly dynamic cardiac environment. In this article, a comprehensive technical overview of continuum robotic catheter systems developed specifically for transcatheter mitral valve interventions is presented. Fundamental design principles of flexible, tendon-driven architectures are examined, highlighting their capacity to navigate tortuous vascular pathways and offer multi-degree-of-freedom control. The integration of advanced sensing technologies, real-time imaging methods, and intelligent control strategies is discussed. Clinical studies and in vivo validations are reviewed, underscoring critical performance metrics such as positional accuracy, procedural safety, and device miniaturization. Persistent challenges are also addressed, including limited high-fidelity data for machine learning, a lack of robust haptic feedback in delicate cardiac tissue manipulation, and regulatory hurdles for complex robotic platforms. Furthermore, emerging innovations in materials science, three-dimensional printing, and sensor fusion are explored, illustrating the potential for next-generation systems that enhance precision while reducing operator workload. Finally, key opportunities for future research are outlined, with an emphasis on personalized navigation algorithms, standardized evaluation protocols, and broader applicability in cardiovascular and endovascular procedures.
{"title":"Continuum Robotic Catheter Systems for Transcatheter Mitral Valve Procedures: A Technical Review","authors":"Di Ding;Tianliang Yao;Haoyu Wang;Xusen Sun;Rong Luo","doi":"10.1109/ACCESS.2025.3548273","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3548273","url":null,"abstract":"Transcatheter mitral valve procedures have transformed the treatment of mitral regurgitation and stenosis by providing less invasive alternatives to conventional open-heart surgery. However, they introduce stringent requirements for catheter navigation, force modulation, and real-time imaging in a highly dynamic cardiac environment. In this article, a comprehensive technical overview of continuum robotic catheter systems developed specifically for transcatheter mitral valve interventions is presented. Fundamental design principles of flexible, tendon-driven architectures are examined, highlighting their capacity to navigate tortuous vascular pathways and offer multi-degree-of-freedom control. The integration of advanced sensing technologies, real-time imaging methods, and intelligent control strategies is discussed. Clinical studies and in vivo validations are reviewed, underscoring critical performance metrics such as positional accuracy, procedural safety, and device miniaturization. Persistent challenges are also addressed, including limited high-fidelity data for machine learning, a lack of robust haptic feedback in delicate cardiac tissue manipulation, and regulatory hurdles for complex robotic platforms. Furthermore, emerging innovations in materials science, three-dimensional printing, and sensor fusion are explored, illustrating the potential for next-generation systems that enhance precision while reducing operator workload. Finally, key opportunities for future research are outlined, with an emphasis on personalized navigation algorithms, standardized evaluation protocols, and broader applicability in cardiovascular and endovascular procedures.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"43275-43288"},"PeriodicalIF":3.4,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10910200","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143621860","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：