检索结果-内蒙古大学图书馆

Innovation in Technology (ASIANCON), Asian Conference on

作者： Buddhin Saroj Samarasinghe K.B.N. Lakmali Dept. of Computer Science Informatics Institute of Technology Colombo 06 Sri Lanka Department of Computer Science Informatics Institute of Technology Colombo 06 Sri Lanka

In recent years snake and insect attacks have become a huge problem worldwide. Most species have similar colors and shapes, which makes it hard to tell them apart using typical techniques. Similarly, identifying different species of bees and wasps can be challenging, especially for non-experts in the field. Therefore, developing a reliable and effective method for recognizing these animals is essential to reducing the issues caused by snake and insect attacks and supporting wildlife conservation efforts. The proposed method in this research utilizes deep ensemble learning and transfer learning techniques to identify snakes, wasps, and bees accurately. Deep ensemble learning involves combining several machine learning models to make better predictions, and transfer learning is a technique that allows pre-trained models to be re-used for new tasks, saving time and computational resources. The proposed method utilizes deep ensemble learning and uses three base transfer learning models, DenseNet201, MobileN etV2, and InceptionResN etV2, for classification. The study obtained a training accuracy of 93%. The research classified 11 types of snakes, four types of bees, and four types of wasps, including all 19 classes. Moreover, the author developed a mobile application for user interaction and utilized Flask API for the logic tier. The mobile application enables users to take pictures or upload images of a snake, wasp, or bee and identify the species accurately. This feature makes the proposed method more accessible to people who encounter these animals daily. Overall, this research represents a promising step towards developing an accurate and effective system for identifying snakes, wasps, and bees, with practical implications for human safety and biodiversity conservation.

关键词：

来源：评论

学校读者我要写书评

暂无评论

AIM-HKR: AI-Driven Molecular Retrosynthesis Using Heterogeneous Knowledge Representations

AIM-HKR: AI-Driven Molecular Retrosynthesis Using Heterogene...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Bomgni, Alain Bertrand Ceskoutsé, Ribot Fleury Téné Njingang, Kevin Jordan Njike Bouetou, Thomas Bouetou Gadhamshetty, Venkataramana Etienne Gnimpieba, Z. Oglala Lakota College KyleSD United States University of Yaounde 1 Ecole Nationale Supérieure Polytechnique Yaounde Cameroon University of Dschang Dept. of Computer Science Dschang Cameroon South Dakota School of Mines and Technology Dept. of Civil and Environmental Engineering Rapid CitySD United States University of South Dakota Dept. of Biomedical Engineering VermillionSD United States

ISBN: (纸本)9798350386226

The synthesis of small molecules is a crucial task across multiple scientific domains, including drug discovery, materials science, and sustainable chemistry. As AI and Machine Learning (ML) continue to advance, these technologies offer transformative potential in molecular synthesis, enhancing efficiency and expanding synthetic possibilities. In particular, small molecule synthesis has profound implications for creating novel therapeutic agents, high-performance materials, and greener chemical processes, but a major challenge remains in designing efficient synthetic routes for target molecules. Retrosynthesis, the process of mapping out synthetic pathways by working backwards from a target molecule, represents a vital step;however, traditional retrosynthesis methods often struggle to predict complex or novel transformations. To address this limitation, we introduce AIM-HKR: AI-Driven Molecular Retrosynthesis Using Heterogeneous Knowledge Representations, a model that leverages graph neural network techniques to enhance retrosynthetic predictions. AIM-HKR integrates information from heterogeneous knowledge graphs, capturing the intricate relationships and analogical reasoning required for retrosynthesis. This model generates type-specific embeddings that reflect both network topology and semantic connections across different entity types, such as molecules, reactions, and functional groups. AIM-HKR's unique capacity to leverage heterogeneous graphs allows it to propose synthetic pathways that extend beyond established precedents, enabling predictions of chemically feasible but previously uncharted transformations. We believe AIM-HKR has the potential to significantly advance molecular synthesis through AI-driven retrosynthesis, establishing a new paradigm in AI-assisted chemistry. © 2024 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Mind the Gap: Evaluating Patch Embeddings from General-Purpose and Histopathology Foundation Models for Cell Segmentation and Classification

arXiv

引用

arXiv 2025年

作者： Vadori, Valentina Peruffo, Antonella Graïc, Jean-Marie Finos, Livio Grisan, Enrico Dept. of Computer Science and Informatics London South Bank University London United Kingdom Dept. of Comparative Biomedicine and Food Science University of Padova Padova Italy Dept. of Statistical Sciences University of Padova Padova Italy

Recent advancements in foundation models have transformed computer vision, driving significant performance improvements across diverse domains, including digital histopathology. However, the advantages of domain-specific histopathology foundation models over general-purpose models for specialized tasks such as cell analysis remain underexplored. This study investigates the representation learning gap between these two categories by analyzing multi-level patch embeddings applied to cell instance segmentation and classification. We implement an encoder-decoder architecture with a consistent decoder and various encoders. These include convolutional, vision transformer (ViT), and hybrid encoders pre-trained on ImageNet-22K or LVD-142M, representing general-purpose foundation models. These are compared against ViT encoders from the recently released UNI, Virchow2, and Prov-GigaPath foundation models, trained on patches extracted from hundreds of thousands of histopathology whole-slide images. The decoder integrates patch embeddings from different encoder dept.s via skip connections to generate semantic and distance maps. These maps are then post-processed to create instance segmentation masks—where each label corresponds to an individual cell—and to perform cell-type classification. All encoders remain frozen during training to assess their pre-trained feature extraction capabilities. Using the PanNuke and CoNIC histopathology datasets, and the newly introduced Nissl-stained CytoDArk0 dataset for brain cytoarchitecture studies, we evaluate instance-level detection, segmentation, and cell-type classification accuracy. This study provides insights into the comparative performance of general-purpose vs. histopathology foundation models, offering guidance for model selection in cell-focused histopathology and brain cytoarchitecture analysis workflows. © 2025, CC BY-NC-SA.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

DeepFuzzLoc Positioning: A Unified Fusion of Fuzzy Clustering and Deep Learning for Scalable Indoor Localization Using Wi-Fi RSSI

DeepFuzzLoc Positioning: A Unified Fusion of Fuzzy Clusterin...

引用

International Symposium on Soft Computing and Intelligent Systems (SCIS)

作者： Muhammad Izzuddin Mahali Jenq-Shiou Leu Nur Achmad Sulistyo Putro Setya Widyawan Prakosa Cries Avian Dept. of Electronic and Computer Engineering National Taiwan University of Science and Technology Taipei Taiwan Dept. of Electronics and Informatics Engineering Education Universitas Negeri Yogyakarta Yogyakarta Indonesia Dept. of Computer Science and Electronics Universitas Gadjah Mada Yogyakarta Indonesia

ISBN: (数字)9798350373332

ISBN: (纸本)9798350373349

This research introduces a novel method for indoor positioning by integrating Fuzzy C-Means clustering with deep learning autoencoders, leveraging Wi-Fi RSSI fingerprinting for precise location prediction and building and floor level classification within a unified framework. The approach combines the comprehensibility of Fuzzy C-Means (FCM) with the robust feature extraction capabilities of deep autoencoders through a dual-phase strategy: extensive offline training utilizing RSSI fingerprint data followed by real-time online position prediction. The model's performance was rigorously tested using 5-Fold cross-validation, examining its effectiveness both with and without the integration of FCM on various cluster numbers. Evaluation metrics, including Euclidean distance error for regression and accuracy, F1-score, precision, and specificity for classification effectiveness, were used for comparative analysis. The results were visualized through regression plots, confusion matrices, and learning curves, supplemented by t-SNE visualizations that provide profound insights into the feature space. These detailed analyses highlight the considerable promise of this integrated approach in enhancing the accuracy and reliability of indoor positioning systems.

关键词： Location awareness Deep learning Training Measurement Accuracy Clustering algorithms Fingerprint recognition Real-time systems Reliability Wireless fidelity

来源：评论

学校读者我要写书评

暂无评论

Image Classification of Decapterus Macarellus Using Ridge Regression 8

Image Classification of Decapterus Macarellus Using Ridge Re...

引用

8th International Conference on Education and Technology, ICET 2022

作者： Jerandu, Charmelia Yunizar Mondolang, Alicia Herlin Sianturi, Corazon Olivia Sianturi, Shine Crossifixio Ximenes, Jenita Felixia Gumelar, Agustinus Bimo Sinlae, Alfry Aristo Jansen Batarius, Patrisius Yuhana, Umi Laili Sooai, Adri Gabriel Universitas Katolik Widya Mandira Dept. of Computer Science Kupang Indonesia Universitas Negeri Malang Dept. of Mathematics Malang Indonesia Universitas Sanata Dharma Dept. of Mathematics Yogyakarta Indonesia Universitas Sanata Dharma Dept. of Computer Science Yogyakarta Indonesia Institut Teknologi Sepuluh Nopember Dept. of Electrical Engineering Surabaya Indonesia Institut Teknologi Sepuluh Nopember Dept. of Informatics Engineering Surabaya Indonesia

ISBN: (数字)9781665492140

ISBN: (纸本)9781665492140

Pelagic fish such as mackerel are a source of protein in Indonesia. However, there is no decapterus macarellus as an open dataset for image processing using various classification algorithms. Where its use includes the sensor-assisted sorting process in checking fresh fish and rotten fish. For this reason, this study aims to provide a classification model for pelagic fish and their primary datasets which is available for free on the IEEE data port. Artificial intelligence is used in the process of guided classification with the help of ground truth for the preparation of fish classes. The dataset used is a primary dataset consisting of fish images, arranged in two classes, namely 71 fresh fish and 96 rotten fish. The methods used are k-NN classifiers, naive bayes and ridge regression. Experiments were drawn up to classify rotten fish and fresh fish. The preprocessing was assisted by InceptionV3 as a feature extraction method. Furthermore, the image data is trained in a ratio of 60:40 for training and testing data. Validation was performed using 2-fold cross validation with the results obtained being 99.4%, 94% and 100% accuracy using the classifiers namely k-NN, naive bayes and ridge regression, respectively. © 2022 IEEE.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Indonesian Hate Speech and Abusive Tweets Classification with Deep Learning Pre-trained Language Models

Indonesian Hate Speech and Abusive Tweets Classification wit...

引用

International Conference of computer and informatics Engineering (IC2IE)

作者： Bagus Tri Yulianto Darmawan Bassamtiano Renaufalgi Irnawan Yoshimi Suzuki Dept. of Computer and Informatics Engineering Politeknik Negeri Jakarta Depok Indonesia Dept. of Computer Science and Engineering University of Yamanashi Kofu Japan

Hate speech and abusive language on social media can spread massively and escalate into conflict between two or more individuals or parties. Previous studies about multi-label classification to identify whether the tweet contains hate speech, abusive language, or not have been conducted with machine learning techniques. They applied machine learning with a combination of feature extraction and achieved a good result in recognizing the tweet with hate speech content. In this research, we apply two approaches: first, we use vanilla pre-trained models of IndoBERT, IndoBERTweet, and Indonesian RoBERTa; second, we combine the three previous models with CNN. Our experiment shows that our proposed method has better results compared to previous research with the best performance achieved by IndoBERTweet+CNN with 93.9% Accuracy.

关键词：

来源：评论

学校读者我要写书评

暂无评论

SE-Half-UNet: Accurate and Low-Cost Retinal Vessel Segmentation from Fundus Images

SE-Half-UNet: Accurate and Low-Cost Retinal Vessel Segmentat...

引用

International Conference on Pattern Analysis and Intelligent Systems (PAIS)

作者： Anouar Khaldi Belal Khaldi Mohamed Ben Bezziane Khaled Hebbache Samir Guediri Siham Hasan Dept. Computer Science University of Kasdi Merbah Ouargla Algeria LIAP Laboratory University of El Oued El Oued Algeria Dept. Computer Science and Informatics University of De Montfort Leicester United Kingdom

ISBN: (数字)9798350350265

ISBN: (纸本)9798350350272

Accurate retinal vessel segmentation from eye fundus images is a major challenge in automated methods. U-Net has been widely used in medical image segmentation, and its effectiveness has led researchers to propose various U-Net variants. However, most of these variants are resource-intensive, suggesting that further development is needed. This study introduces a novel approach called SE-Half-UNet, which derives from the Half-UNet framework by incorporating the Squeeze-and-Excitation (SE) Module to bolster the network’s representative capacity by emphasizing channel relationships within feature maps. Our method achieves superior performance in retinal vessel segmentation, with a high F1-Score of 0.8303 and 0.8372 for DRIVE and CHASE_DB datasets, respectively. This new method facilitates the accurate analysis of eye images and encourages computer research advances.

关键词： Image segmentation Channel capacity Semantics Retinal vessels Pattern analysis Intelligent systems biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

A Call Admission Control Mechanism based both on Queueing and the Probabilistic Bandwidth Reservation Policy in Mobile Hotspots

A Call Admission Control Mechanism based both on Queueing an...

引用

2022 Panhellenic Conference on Electronics and Telecommunications, PACET 2022

作者： Vlasakis, Marinos Keramidi, Irene Moscholios, Ioannis Sarigiannidis, Panagiotis Logothetis, Michael Hellenic Open University School of Science and Technology Patras Greece University of Peloponnese Dept. Informatics and Telecommunications Tripolis Greece University of W. Macedonia Dept. Electrical and Computer Engineering Kozani Greece Dept. Electrical and Computer Engineering University of Patras Patras Greece

ISBN: (数字)9798350399585

ISBN: (纸本)9798350399585

In this paper we study a mobility-aware call admission control mechanism applied in a mobile hotspot. We consider a vehicle, which can alternate between stop and moving phases and has an access point of a fixed capacity. In the stop phase, the vehicle accommodates both new and handover calls. To prioritize handover calls a probabilistic bandwidth reservation (BR) policy is applied where some of the system's resources are reserved for handover calls. Based on this policy, new calls may enter the reservation space with a predefined probability. In addition, handover calls can wait in a queue of finite size if there are no available resources at the time of their arrival. In the moving phase, the vehicle services only new calls under the probabilistic BR policy. In both phases, arriving calls follow a Poisson process, require a single bandwidth unit from the system and have an exponentially distributed service time. To analytically determine the most significant performance measures such as call blocking probabilities an efficient iterative algorithm is presented. © 2022 IEEE.

关键词： Bandwidth

来源：评论

学校读者我要写书评

暂无评论

Robust Self-Guided Deep Image Prior 48

Robust Self-Guided Deep Image Prior

引用

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Bell, Evan Liang, Shijun Qu, Qing Ravishankar, Saiprasad Michigan State University Dept. of Mathematics East LansingMI United States University of Michigan Dept. of Electrical and Computer Engineering Ann ArborMI United States Michigan State University Dept. of Biomedical Engineering East LansingMI United States Michigan State University Dept. of Computational Mathematics Science and Engineering East LansingMI United States

ISBN: (纸本)9781728163277

In this work, we study the deep image prior (DIP) for reconstruction problems in magnetic resonance imaging (MRI). DIP has become a popular approach for image reconstruction, where it recovers the clear image by fitting an overparameterized convolutional neural network (CNN) to the corrupted/undersampled measurements. To improve the performance of DIP, recent work shows that using a reference image as an input often leads to improved reconstruction results compared to vanilla DIP with random input. However, obtaining the reference input image often requires supervision and hence is difficult in practice. In this work, we propose a self-guided reconstruction scheme that uses no training data other than the set of undersampled measurements to simultaneously estimate the network weights and input (reference). We introduce a new regularization that aids the joint estimation by requiring the CNN to act as a powerful denoiser. The proposed self-guided method gives significantly improved image reconstructions for MRI with limited measurements compared to the conventional DIP and the reference-guided method while eliminating the need for any additional data. © 2023 IEEE.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Classification of Functional and Nonfunctional Hand Movement Using Deep Learning and First-Person View Video

Classification of Functional and Nonfunctional Hand Movement...

引用

2023 Congress in computer science, computer Engineering, and Applied Computing, CSCE 2023

作者： Tran, Tan Chang, Lin-Ching Lum, Peter S. Dept. of Electrical Engineering and Computer Science The Catholic University of America WashingtonDC20064 United States The Catholic University of America Dept. of Biomedical Engineering WashingtonDC20064 United States MedStar National Rehabilitation Network WashingtonDC20064 United States

ISBN: (纸本)9798350327595

Upper extremity (UE) rehabilitation of motor function is needed post-stroke. Although UE rehabilitation is widely used in practice, it is difficult to evaluate the effectiveness of UE treatments. There are various approaches to evaluate UE motor function such as performing tasks in the clinician's office, patient's self-report, or wearable devices equipped with accelerometer/gyro sensors to analyze the data captured from stroke survivors. Our previous study reported that using a Random Forest machine learning algorithm with accelerometry sensor data obtained by single wrist-worn sensor, it is possible to differentiate between functional and non-functional UE movements in stroke survivors with good accuracy. However, to build the model, a third-person view video was captured and manually annotated by experts to provide ground truth labels for training. This task is very time-consuming and subjective. In this study, we propose a framework that uses deep learning (DL) to classify functional and non-functional hand movement in videos captured from a wearable camera (i.e., first-person view videos). The proposed system is fully automated and consists of two DL networks. The first DL network performs hand pose estimation from a 2-second sequence of frames which extract pre-defined key points on the hands. The second DL network then takes these key points as input and classifies the sequence of frames to either functional or non-functional movement. The proposed system has two major impacts on rehabilitation. First, it can provide initial annotations on the frames of videos that can significantly reduce the time required for human annotations. Second, analyzing the effectiveness of UE rehabilitation using the first-person view videos enables us to measure the outcomes of UE treatments in stroke survivors' homes. In addition, the system can deliver timely feedback to stroke survivors that has the potential to increase the use of the affected UE in the community. © 2023 IEEE

关键词： deep learning - Regular Research Paper rehabilitation stroke upper extremity

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：