检索结果-内蒙古大学图书馆

Pre-trained SAM as data augmentation for image segmentation

CAAI Transactions on Intelligence technology 2025年第1期10卷 268-282页

作者： Junjun Wu Yunbo Rao Shaoning Zeng Bob Zhang Yangtze Delta Region Institute(Huzhou) University of Electronic Science and Technology of ChinaHuzhouChina School of Information and Software Engineering University of Electronic Science and Technology of ChinaChengduChina Pattern Analysis and Machine Intelligence Research Group Department of Computer and Information ScienceUniversity of MacaoMacaoChina

Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the ***,data augmentation mainly involved some simple transformations of ***,in order to increase the diversity and complexity of data,more advanced methods appeared and evolved to sophisticated generative ***,these methods required a mass of computation of training or *** this paper,a novel training-free method that utilises the Pre-Trained Segment Anything Model(SAM)model as a data augmentation tool(PTSAM-DA)is proposed to generate the augmented annotations for *** the need for training,it obtains prompt boxes from the original annotations and then feeds the boxes to the pre-trained SAM to generate diverse and improved *** this way,annotations are augmented more ingenious than simple manipulations without incurring huge computation for training a data augmentation *** comparative experiments on three datasets are conducted,including an in-house dataset,ADE20K and *** this in-house dataset,namely Agricultural Plot Segmentation Dataset,maximum improvements of 3.77%and 8.92%are gained in two mainstream metrics,mIoU and mAcc,***,large vision models like SAM are proven to be promising not only in image segmentation but also in data augmentation.

关键词： data augmentation image segmentation large model segment anything model

来源：评论

学校读者我要写书评

暂无评论

Adapting Pedagogical Frameworks to Online Learning: Simulating the Tutor Role with LLMs and the QUILT Model 6

Adapting Pedagogical Frameworks to Online Learning: Simulati...

引用

6th International Conference on Advancements in Computing, ICAC 2024

作者： Madhubhashana, Sandaru Ranhinda, Vibhashana Jayakody, Dineth Shavinda, Lakmuthu Nawarathne, Malithi Haddela, Prasanna Sri Lanka Institute of Information Technology Dept. of Software Engineering Malabe Sri Lanka Sri Lanka Institute of Information Technology Faculty of Computing Malabe Sri Lanka

ISBN: (纸本)9798331517878

This paper presents SoloScholar, a novel asynchronous online learning platform designed to deliver personalized education to large undergraduate student groups by integrating established pedagogical frameworks with Large Language Models (LLMs). Leveraging Bloom's Revised Taxonomy, learners are categorized into beginner, intermediate, and advanced levels to personalize the educational content. The platform employs the Questioning and Understanding to Improve Learning and Thinking (QUILT) model to structure tutorial sessions and simulate the teaching role of a tutor at a Higher Education institute (HEI). To generate contextually relevant and cognitively appropriate questions, carefully crafted prompts are fed into LLMs, synthesizing information from reliable knowledge sources. The efficacy of SoloScholar was evaluated through a dual assessment approach: a technical evaluation using the Retrieval Augmented Generation Assessment (RAGAS) framework and a human evaluation involving academic experts. The technical evaluation demonstrated high scores in answer relevancy and context utilization across all learning levels. Similarly, quantitative responses from academic experts indicated strong satisfaction with the quality and appropriateness of the generated content. The findings underscore the potential of integrating LLMs with pedagogical models to enhance personalization in online education. SoloScholar effectively addresses challenges associated with large-group teaching by providing adaptive learning experiences that cater to diverse cognitive needs. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Deep Learning to Classify Bacterial Species in the same Genus

Deep Learning to Classify Bacterial Species in the same Genu...

引用

2024 International Conference on Electronics, information, and Communication, ICEIC 2024

作者： Sheela, Sherin Piang, May Phu Sakorntanant, Sakda Poomrittigul, Suvit Treebupachatsakul, Treesukon School of Engineering King Mongkut's Institute of Technology Ladkrabang Department of Biomedical Engineering Thailand Pathumwan Institute of Technology Faculty of Science and Technology Department of Software Engineering and Information System Bangkok Thailand School of Information Technology King Mongkut's Institute of Technology Ladkrabang Thailand

ISBN: (纸本)9798350371888

Bacterial strains in the same genus share highly similar morphology, gram-staining characteristics, colony sizes, and spatial arrangements. Therefore, identifying them by deep learning can be quite challenging. This study aimed to assess the classification of 7 species of bacteria from 2 genera of Bacillus and Vibrio by using 8 Convolutional Neural Network (CNN) models. We implemented Python programming along with Keras API within the Jupyter Notebook. The models were constructed and evaluated under unbalanced and balanced datasets by augmentation (rotation, flip, etc.). Transfer learning with fine-tuning, and pre-processing of mixup and label smoothing were also applied to reduce overfitting and enhance generalization. Based on the experimental results on private dataset, the results of InceptionResNetV2 emerged as the top-performing model with a notable accuracy of 82.8%, 88.6% precision, 78.4% recall, and 78.0% F1-score when label smoothing was applied at 0.5 on balanced dataset. © 2024 IEEE.

关键词： Neural network models

来源：评论

学校读者我要写书评

暂无评论

A Multimodal Interaction System for Speech-Based Autism Intervention in Sinhala-Speaking Sri Lankan Children using the NAO Robot 6

A Multimodal Interaction System for Speech-Based Autism Inte...

引用

6th International Conference on Advancements in Computing, ICAC 2024

作者： Jayawardena, Aparna Manukalpa, Chan Sri Bopage, Heshani Panduwawala, Poorna Pulasinghe, Koliya Rajapakshe, Samantha Sri Lanka Institute of Information Technology Department of Software Engineering Malabe Sri Lanka Sri Lanka Institute of Information Technology Department of Information Technology Malabe Sri Lanka Sri Lanka Institute of Information Technology Department of Computer Science Malabe Sri Lanka

ISBN: (纸本)9798331517878

This paper presents a multimodal interaction system designed for speech-based autism intervention in Sinhala-speaking Sri Lankan children, utilizing the NAO robot. The system integrates four components, language content detection, dialog management, voice synthesis, and gesture synthesis, to create an engaging and adaptive therapeutic environment. By leveraging these technologies, the system provides real-time, personalized responses that align with the cultural and linguistic needs of Sinhala-speaking children with autism spectrum disorder (ASD). The NAO robot facilitates natural and supportive interaction through synchronized speech and gestures, aiming to improve communication and social skills in children with ASD. Preliminary evaluations demonstrate the potential of the system to enhance engagement and responsiveness in therapy sessions, marking a significant step forward in developing specialized interventions for autism in non-Western, non-English-speaking contexts. © 2024 IEEE.

关键词： Speech synthesis

来源：评论

学校读者我要写书评

暂无评论

Handwritten Prescription Recognition Using VGG Based Architecture with Bi-LSTM

Handwritten Prescription Recognition Using VGG Based Archite...

引用

2024 International Research Conference on Smart Computing and Systems engineering, SCSE 2024

作者： Kavinda, Isuru Fernando, Harinda Sri Lanka Institute of Information Technology Department of Computer Science and Software Engineering Malabe Sri Lanka Sri Lanka Institute of Information Technology Department of Computer System Engineering Malabe Sri Lanka

ISBN: (纸本)9798350375688

Illegible handwriting on medical prescriptions poses a significant challenge, often leading to the misinterpretation of drug names and dosages. This issue primarily stems from doctors' use of Latin abbreviations, driven by time constraints during patient consultations. The resultant discrepancies in prescription interpretation can potentially expose patients to incorrect medications and dosages, leading to adverse effects. In response to this concern, this study proposes a novel system integrating Convolution Neural Network (CNN) and Recurrent Neural Network (RNN) models. Promising findings are reported in this study, achieving a loss of 0.4874 and an accuracy of 0.83 in the latest trial. The final model offers a potential solution to the persistent problem of illegible prescriptions, enhancing patient safety in the healthcare domain. © 2024 IEEE.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Multi-Objective Equilibrium Optimizer for Feature Selection in High-Dimensional English Speech Emotion Recognition

引用

Computers, Materials & Continua 2024年第2期78卷 1957-1975页

作者： Liya Yue Pei Hu Shu-Chuan Chu Jeng-Shyang Pan Fanli Business School Nanyang Institute of TechnologyNanyang473004China School of Computer and Software Nanyang Institute of TechnologyNanyang473004China College of Computer Science and Engineering Shandong University of Science and TechnologyQingdao266590China Department of Information Management Chaoyang University of TechnologyTaichung413310Taiwan

Speech emotion recognition(SER)uses acoustic analysis to find features for emotion recognition and examines variations in voice that are caused by *** number of features acquired with acoustic analysis is extremely high,so we introduce a hybrid filter-wrapper feature selection algorithm based on an improved equilibrium optimizer for constructing an emotion recognition *** proposed algorithm implements multi-objective emotion recognition with the minimum number of selected features and maximum ***,we use the information gain and Fisher Score to sort the features extracted from ***,we employ a multi-objective ranking method to evaluate these features and assign different importance to *** with high rankings have a large probability of being ***,we propose a repair strategy to address the problem of duplicate solutions in multi-objective feature selection,which can improve the diversity of solutions and avoid falling into local *** random forest and K-nearest neighbor classifiers,four English speech emotion datasets are employed to test the proposed algorithm(MBEO)as well as other multi-objective emotion identification *** results illustrate that it performs well in inverted generational distance,hypervolume,Pareto solutions,and execution time,and MBEO is appropriate for high-dimensional English SER.

关键词： Speech emotion recognition filter-wrapper high-dimensional feature selection equilibrium optimizer multi-objective

来源：评论

学校读者我要写书评

暂无评论

Sub-10-nm Tunnel Field-Effect Transistor with Schottky Drain

Sub-10-nm Tunnel Field-Effect Transistor with Schottky Drain

引用

2023 IEEE International Students' Conference on Electrical, Electronics and Computer Science, SCEECS 2023

作者： Liu, Ying Wang, Yuan Shan, Chan Ou-Yang, Lian-Ying Wang, Shu-Rong Ruan, Ping-Ting Xu, Ge-Jing Xiamen Institute of Software Technology Department of Software Technology Xiamen China College of Ocean Information Engineering Jimei University Xiamen China Xiamen Institute of Software Technology Department of Economics and Management Xiamen China College of Software Quanzhou University of Information Engineering Quanzhou China

ISBN: (纸本)9798350398748

An SD-TFET, which is a tunnel field effect transistor with Schottky drain, is proposed in this paper for sub-10-nm applications. On-off switching characteristics were examined by examining some of the critical device parameters, including the metal work-function in a Schottky drain. When designing SD- TFETs for low-power integrated circuits, a moderate Schottky drain work-function is preferred. © 2023 IEEE.

关键词： Silicides

来源：评论

学校读者我要写书评

暂无评论

YGC-SLAM:A visual SLAM based on improved YOLOv5 and geometric constraints for dynamic indoor environments

引用

虚拟现实与智能硬件(中英文) 2025年第1期7卷 62-82页

作者： Juncheng ZHANG Fuyang KE Qinqin TANG Wenming YU Ming ZHANG Department of Automation University of Information Science&TechnologyNanjing 210044China Department of Software University of Information Science&TechnologyNanjing 210044China Wuxi Research Institute of Nanjing University of Information Engineering Wuxi 214101China Department of Rail Transport University of WuxiWuxi 214101China

Background As visual simultaneous localization and mapping(SLAM)is primarily based on the assumption of a static scene,the presence of dynamic objects in the frame causes problems such as a deterioration of system robustness and inaccurate position *** this study,we propose a YGC-SLAM for indoor dynamic environments based on the ORB-SLAM2 framework combined with semantic and geometric constraints to improve the positioning accuracy and robustness of the *** First,the recognition accuracy of YOLOv5 was improved by introducing the convolution block attention model and the improved EIOU loss function,whereby the prediction frame converges quickly for better *** improved YOLOv5 was then added to the tracking thread for dynamic target detection to eliminate dynamic ***,multi-view geometric constraints were used for re-judging to further eliminate dynamic points while enabling more useful feature points to be retained and preventing the semantic approach from over-eliminating feature points,causing a failure of map *** K-means clustering algorithm was used to accelerate this process and quickly calculate and determine the motion state of each cluster of pixel ***,a strategy for drawing keyframes with de-redundancy was implemented to construct a clear 3D dense static point-cloud *** Through testing on TUM dataset and a real environment,the experimental results show that our algorithm reduces the absolute trajectory error by 98.22%and the relative trajectory error by 97.98%compared with the original ORBSLAM2,which is more accurate and has better real-time performance than similar algorithms,such as DynaSLAM and *** The YGC-SLAM proposed in this study can effectively eliminate the adverse effects of dynamic objects,and the system can better complete positioning and map building tasks in complex environments.

关键词： Visual SLAM Dynamic SLAM Target detection Geometric constraints

来源：评论

学校读者我要写书评

暂无评论

Mutli-focus image fusion based on guided filter and image matting network

引用

Multimedia Tools and Applications 2025年第3期84卷 1239-1259页

作者： Zhu, Puchao Li, Xue Wang, Puming Jin, Xin Yao, Shaowen School of Software Yunnan University East Outer Ring Road Chenggong District Yunnan Kunming650091 China School of Electronic Information Engineering Henan Institute of Technology Xinxiang453003 China

The problem of limited depth-of-field is one of the major disadvantages in optical imaging devices, whereas multifocus image fusion(MFIF), as an enhancement technique, can effectively solve this situation. The purpose of MFIF method is to fuse the different focus regions from the multifocus image pair so as to obtain a full-clear fusion result. However, the inaccurate decision maps in fusion methods often leads to blurred object boundary or artifact. To solve this problem, we propose a novel method combining guided filter and deep image matting network for multifocus image fusion tasks. The first step is to utilize guided filter to generate high-frequency information of multifocus image pairs, as to get the initial decision maps based on pixel-level maximum comparison. Aiming for the better fusion images, these decision maps need to be further refined. Then, the trimaps corresponding to the initial decision images are gained via the basic morphological operations, and we utilize the deep image matting network to refine the uncertain regions of trimaps, which can effectively reduce the misclassified pixels and yield the final decision maps with less noise. Finally, the fusion results are constructed by using the final decision maps. This approach effectively distinguishes between focused and non-focused areas, preserving the complete and sharp contour information of objects in the fused images. The experiments illustrate the superior visual effect and objective evaluation on public datasets are achieved by the designed fusion method compared with the advanced fusion methods. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

A Comparison of Deep Learning and Machine Learning Methods for Water Quality Prediction 3

A Comparison of Deep Learning and Machine Learning Methods f...

引用

3rd International Conference on Intelligent Data Communication Technologies and Internet of Things, IDCIoT 2025

作者： Srinivasa Gupta, N. Valarmathi, B. Hiwanj, Ishaan Department of Manufacturing Engineering School of Mechanical Engineering Vellore Institute of Technology Tamil Nadu Vellore India Department of Software and Systems Engineering School of Computer Science Engineering and Information Systems Vellore Institute of Technology Tamil Nadu Vellore India Department of Information Technology School of Computer Science Engineering and Information Systems Vellore Institute of Technology Tamil Nadu Vellore India

ISBN: (纸本)9798331527549

Predicting water quality is essential to preserving human health and environmental sustainability. Traditional water quality assessment methods often face scalability and real-time monitoring limitations. With accuracies of 62%, 72%, 83%, 69%, 63%, 66%, 71%, 63%, and 64%, respectively, the current techniques utilized were Logistic Regression, Decision Trees, Random Forest Regressor, Extreme Gradient Boosting, Naive Bayes, K-nearest neighbors, Support Vector Machine, AdaBoost, and Bagging [9]. This study addresses these challenges by leveraging Adaptive Synthetic Sampling (ADASYN) to balance the dataset and evaluating model performance on datasets of 5,000 and 10,000 entries per class. A robust dataset obtained from Kaggle was used, with five models – Long Short-Term Memory (LSTM), Feed Forward Neural Network (FFNN), Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), and Random Forest – evaluated and compared. The proposed methods demonstrate significant improvements in accuracy, with XGBoost achieving the highest accuracy of 95.53%, followed by Random Forest at 93.98%. This work underscores the importance of advanced machine learning techniques in addressing the limitations of traditional methods, enhancing accuracy, scalability, and adaptability in water quality prediction. These findings contribute to advancing environmental monitoring and management practices with reliable, data-driven insights. © 2025 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：