检索结果-内蒙古大学图书馆

Multimodal contrastive learning for radiology report generation

Journal of Ambient Intelligence and Humanized Computing 2023年第8期14卷 11185-11194页

作者： Wu, Xing Li, Jingwen Wang, Jianjia Qian, Quan School of Computer Engineering and Science Shanghai University Shanghai200444 China Shanghai Institute for Advanced Communication and Data Science Shanghai University Shanghai200444 China

Automated radiology report generation can not only lighten the workload of clinicians but also improve the efficiency of disease diagnosis. However, it is a challenging task to generate semantically coherent radiology reports that are also highly consistent with medical images. To meet the challenge, we propose a Multimodal Recursive model with Contrastive Learning (MRCL). The proposed MRCL method incorporates both visual and semantic features to generate "Impression" and "Findings" of radiology reports through a recursive network, in which a contrastive pre-training method is proposed to improve the expressiveness of both visual and textual representations. Extensive experiments and analyses prove the efficacy of the proposed MRCL, which can not only generate semantically coherent radiology reports but also outperform state-of-the-art methods. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Effectiveness of K-means Clustering-Based Outlier Handling Techniques in Colonic Condition Endoscopy Image Classification Using CNN 13

Effectiveness of K-means Clustering-Based Outlier Handling T...

引用

13th IEEE International Conference on Communication, Networks and Satellite, COMNETSAT 2024

作者： Athiyah, Ummi Rustad, Supriadi Soeleman, Moch Arief Muljono Akrom, Muhamad Dian Nuswantoro University Faculty of Computer Science Semarang Indonesia Telkom University Department of Data Science Purwokerto Indonesia

ISBN: (纸本)9798350368086

Colorectal cancer is one of the leading causes of cancer death worldwide, so accurate early detection is needed. Endoscopic imaging technology plays a vital role in diagnosis, but the presence of outliers in endoscopic images can significantly affect the accuracy of the classification model. This study addresses this challenge by applying Histogram of Oriented Gradients (HOG) for feature extraction and K-means Clustering for outlier detection and removal. A dataset of colon endoscopic images was obtained from the public database Gastrovision, which was divided into two classes: normal and abnormal conditions. This study applied a two-step modeling process, first with the original dataset and then with the cleaned dataset after outlier removal. The results showcased a significant improvement in model performance after outlier handling, with accuracy increasing from 61% to 85%. The Area Under Curve (AUC) of the ROC curve also increased from 0.84 to 0.92, indicating a better model ability to distinguish between normal and abnormal conditions. These findings emphasize the importance of data quality and effective outlier management in training a machine-learning model for colon cancer endoscopic image conditions. © 2024 IEEE.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

PrimeComposer: Faster Progressively Combined Diffusion for Image Composition with Attention Steering 24

PrimeComposer: Faster Progressively Combined Diffusion for I...

引用

32nd ACM International Conference on Multimedia, MM 2024

作者： Wang, Yibin Zhang, Weizhong Zheng, Jianwei Jin, Cheng School of Computer Science Fudan University Shanghai China School of Data Science Fudan University Shanghai China College of Computer Science and Technology Zhejiang University of Technology Hangzhou China

ISBN: (纸本)9798400706868

Image composition involves seamlessly integrating given objects into a specific visual context. Current training-free methods rely on composing attention weights from several samplers to guide the generator. However, since these weights are derived from disparate contexts, their combination leads to coherence confusion and loss of appearance information. These issues worsen with their excessive focus on background generation, even when unnecessary in this task. This not only impedes their swift implementation but also compromises foreground generation quality. Moreover, these methods introduce unwanted artifacts in the transition area. In this paper, we formulate image composition as a subject-based local editing task, solely focusing on foreground generation. At each step, the edited foreground is combined with the noisy background to maintain scene consistency. To address the remaining issues, we propose PrimeComposer, a faster training-free diffuser that composites the images by well-designed attention steering across different noise levels. This steering is predominantly achieved by our Correlation Diffuser, utilizing its self-attention layers at each step. Within these layers, the synthesized subject interacts with both the referenced object and background, capturing intricate details and coherent relationships. This prior information is encoded into the attention weights, which are then integrated into the self-attention layers of the generator to guide the synthesis process. Besides, we introduce a Region-constrained Cross-Attention to confine the impact of specific subject-related tokens to desired regions, addressing the unwanted artifacts shown in the prior method thereby further improving the coherence in the transition area. Our method exhibits the fastest inference efficiency and extensive experiments demonstrate our superiority both qualitatively and quantitatively. The code is available at https://***/CodeGoat24/PrimeComposer. © 2024 ACM.

关键词： Diffusers (optical)

来源：评论

学校读者我要写书评

暂无评论

MDNN:Predicting Student Engagement via Gaze Direction and Facial Expression in Collaborative Learning

引用

computer Modeling in Engineering & sciences 2023年第7期136卷 381-401页

作者： Yi Chen Jin Zhou Qianting Gao Jing Gao Wei Zhang School of Computer Science Central China Normal UniversityWuhan430079China Computer Science School of ScienceRensselaer Polytechnic InstituteTroy12180USA National Engineering Laboratory of Educational Big Data Application Technology Central China Normal UniversityWuhan430079China

Prediction of students’engagement in aCollaborative Learning setting is essential to improve the quality of *** learning is a strategy of learning through groups or *** cooperative learning behavior occurs,each student in the group should participate in teaching *** showed that students who are actively involved in a class gain *** behavior and facial expression are important nonverbal indicators to reveal engagement in collaborative learning *** studies require the wearing of sensor devices or eye tracker devices,which have cost barriers and technical interference for daily teaching *** this paper,student engagement is automatically analyzed based on computer *** tackle the problem of engagement in collaborative learning using a multi-modal deep neural network(MDNN).We combined facial expression and gaze direction as two individual components of MDNN to predict engagement levels in collaborative learning *** multi-modal solution was evaluated in a real collaborative *** results show that the model can accurately predict students’performance in the collaborative learning environment.

关键词： Engagement facial expression deep network gaze

来源：评论

学校读者我要写书评

暂无评论

Detecting Cyber Bullying in the Social Media Using Deep Learning Algorithms 15th

Detecting Cyber Bullying in the Social Media Using Deep Lear...

引用

15th International Conference on Soft Computing and Pattern Recognition, SoCPaR 2023 and 14th World Congress on Nature and Biologically Inspired Computing, NaBIC 2023

作者： Mridul, Aunik Hasan Ahsan, Nowreen Sultana, Zakia Asif, Asifuzzaman Computer Science and Engineering Daffodil International University Dhaka Bangladesh Department of Data Science Clarkson University New York United States Computer Science and Engineering Lovely Professional University Punjab India

ISBN: (纸本)9783031810824

Cyberbullying is a pressing issue in the digital age, especially in non-English-speaking communities. This study focuses on using deep learning techniques to detect cyberbullying in the Bangla language. The research employs a Bangla dataset that includes both cyberbullying and non-cyberbullying text, which undergoes preprocessing, tokenization, and sequence transformation for input into a LSTM recurrent neural network. The LSTM model incorporates word embedding and dropout techniques to enhance its performance. Evaluation metrics like accuracy and AUC-ROC score are employed to assess the model's effectiveness. The study investigates many methods, such as vernacular-specific preparation and data augmenting, that are specifically designed for the identification of cyberbullying in the Bangla language. The findings demonstrate deep learning's capacity to identify cyberbullying in Bangla, which makes a significant contribution to resolving this problem in non-English settings. The research reveals impressive accuracy rates, with LSTM achieving 99.99% accuracy, and a different approach, Convolutional Neural Network (CNN), yielding 99.022%. Ensemble models, including Gradient Boosting, XGBoost, and Random Forest, are also employed and optimized using hyperparameter tuning. Notably, LSTM stands out as the most reliable classifier for cyberbullying detection, as confirmed by the experimental findings and recent research. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Random forests

来源：评论

学校读者我要写书评

暂无评论

Graph Classification with GNNs: Optimisation, Representation & Inductive Bias 27

Graph Classification with GNNs: Optimisation, Representation...

引用

27th European Conference on Artificial Intelligence, ECAI 2024

作者： Kumar, P. Krishna Ramaswamy, Harish G. Department of Computer Science & Engineering IIT Madras India Department of Data Science & Artificial Intelligence IIT Madras India

ISBN: (纸本)9781643685489

Theoretical studies on the representation power of GNNs have been centered around understanding the equivalence of GNNs, using WL-Tests for detecting graph isomorphism. In this paper, we argue that such equivalence ignores the accompanying optimization issues and does not provide a holistic view of the GNN learning process. We illustrate these gaps between representation and optimization with examples and experiments. We also explore the existence of an implicit inductive bias (e.g. fully connected networks prefer to learn low frequency functions in their input space) in GNNs, in the context of graph classification tasks. We further prove theoretically that the message-passing layers in the graph, have a tendency to search for either discriminative subgraphs, or a collection of discriminative nodes dispersed across the graph, depending on the different global pooling layers used. We empirically verify this bias through experiments over real-world and synthetic datasets. Finally, we show how our work can help in incorporating domain knowledge via attention based architectures, and can evince their capability to discriminate coherent subgraphs. © 2024 The Authors.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Learning in sinusoidal spaces with physics-informed neural networks

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第3期5卷 985-1000页

作者： Wong, Jian Cheng Ooi, Chin Chun Gupta, Abhishek Ong, Yew Soon Institute of High Performance Computing Agency for Science Technology and Research Singapore138632 Singapore School of Computer Science and Engineering Nanyang Technological University Singapore639798 Singapore Singapore Institute of Manufacturing Technology Agency for Science Technology and Research Singapore138632 Singapore Agency for Science Technology and Research Singapore138632 Singapore Data Science and Artificial Intelligence Research Centre School of Computer Science and Engineering Nanyang Technological University Singapore138632 Singapore

A physics-informed neural network (PINN) uses physics-Augmented loss functions, e.g., incorporating the residual term from governing partial differential equations (PDEs), to ensure its output is consistent with fundamental physics laws. However, it turns out to be difficult to train an accurate PINN model for many problems in practice. In this article, we present a novel perspective of the merits of learning in sinusoidal spaces with PINNs. By analyzing behavior at model initialization, we first show that a PINN of increasing expressiveness induces an initial bias around flat output functions. Notably, this initial solution can be very close to satisfying many physics PDEs, i.e., falling into a localminimum of the PINN loss that onlyminimizes PDE residuals, while still being far from the true solution that jointly minimizes PDE residuals and the initial and/or boundary conditions. It is difficult for gradient descent optimization to escape from such a local minimum trap, often causing the training to stall. We then prove that the sinusoidalmapping of inputs-in an architecture we label as sf-PINN-is effective to increase input gradient variability, thus avoiding being trapped in such deceptive local minimum. The level of variability can be effectively modulated to match high-frequency patterns in the problem at hand. A key facet of this article is the comprehensive empirical study that demonstrates the efficacy of learning in sinusoidal spaces with PINNs for a wide range of forward and inversemodeling problems spanning multiple physics domains. Impact Statement-Falling under the emerging field of physicsinformed machine learning, PINN models have tremendous potential as a unifying AI framework for assimilating physics theory and measurement data. However, they remain infeasible for broad science and engineering applications due to computational cost and training challenges, especially for more complex problems. Instead of focusing on empirical demonstration of appli

关键词： Gradient methods

来源：评论

学校读者我要写书评

暂无评论

Diabetic Retinopathy Detection using Deep Learning Model ResNet15 2

Diabetic Retinopathy Detection using Deep Learning Model Res...

引用

2nd International Conference on Disruptive Technologies, ICDT 2024

作者： Rai, Bipin Kumar Ojha, Harsh Srivastava, Ishani Dayananda Sagar University Department of Computer Science & Engineering Bengaluru India Abes Institute of Technology Department of Computer Science & Engineering- Data Science U.P. Ghaziabad India

ISBN: (纸本)9798350371055

Diabetic Retinopathy stands out as one of the common retinal diseases which has a significant threat to vision and can lead to blindness. Several problems occurred due to DR can be stopped by controlling blood glucose and timely treatment, but the manual assessment of retinal images is often a lengthy process and can prevail human errors. This paper attempts in the direction of finding an automated way to classify the set of retinal images. The study starts with the collection of retinal images and pre- processing them, followed by training of a CNN model, Utilizing Deep Learning methodologies, CNN classifies the degree of diabetic retinopathy (DR) into four categories: 0 (no DR) to 4 (proliferative DR). We evaluate our model's performance using a diverse dataset. We use Convolutional Neural Networks (CNNs) strength to DR detection, which involves three major difficulties: Classification, Segmentation and Detection. We work towards developing a GUI based system to store and maintain a series of predictions with the patient's id and name. Twilio API have been used to make SMS connectivity to patients possible in case they are not accessible. Many research activities in the previous years have many limitations and disadvantages like models used other than Resnet have low accuracy, not capturing intricate features and patterns in images, Vanishing Gradient Descent and low performance for image classification. Resnet models that are previously used in DR detection with Resnet152, we can conclude that it has the lowest Training loss, lowest Test loss and it has the lowest loss Difference. We have worked on a RESNET-152 pre-trained deep learning model, which consists of thousands of layers and allows for the capturing of intricate features, high performance of image classification, and high accuracy, in order to overcome the issues and limitations of earlier developed models. © 2024 IEEE.

关键词： Deep Learning Diabetic Retinopathy (DR) Fundus images Ophthalmologist Proliferative DR Resnet152

来源：评论

学校读者我要写书评

暂无评论

Polyp Detection Using U-Net Neural Network Based Algorithm 2

Polyp Detection Using U-Net Neural Network Based Algorithm

引用

2nd International Conference on Disruptive Technologies, ICDT 2024

作者： Rai, Bipin Kumar Singh, Devansh Shukla, Archita Dayananda Sagar University Department of Computer Science & Engineering Bengaluru India Abes Institute of Technology Department of Computer Science & Engineering-Data Science U.P. Ghaziabad India

ISBN: (纸本)9798350371055

Polyp is an earlier stage of cancer development in gastro-intestinal tract. Despite the fact that numerous techniques for automatic segmentation and detection of polyps have been developed, it still remains an open problem. Examining color images captured by the board camera in endoscopy is contemplated the most reliable method for detectingdeveloping polyps. To automate this process, in this paper we present a modified U-Net Neural Network based algorithm. This algorithm is a binary classifier which labels the pixels as part of the polyp or not. This model can analyze the video of GItract frame by frame and produce the collection of frames with possibility of presence of polyp thus reducing the time required to process the GI-tract. Detecting, localizing, and segmenting polyps using Kvasir-SEG, an open-access dataset of colonoscopy images, we benchmark various modern day approaches in this work, assessing bothmethod speed and accuracy. This is attributed to outstanding performance of image classification compared to preceding techniques. The automatically identifying polyps serves to aid gastroenterologists during colonoscopies. While existing literature contains publications addressing the challenge of polyp detection, many of these systems remain confined toresearch settings and lack implementation for clinical use. Hence, we present the inaugural fully open-source automated polyp detection system, which not only outperforms the best- performing system documented in the literature but is also prepared for immediate clinical application. © 2024 IEEE.

关键词： Colonoscopy ColonSegNet Colorectal cancer Kvasir-SEG Medical image segmentation Polyp Detection U-net

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Survey on Deep Learning for Enhanced Node Position Prediction in Vehicular Ad-Hoc Networks 4

A Comprehensive Survey on Deep Learning for Enhanced Node Po...

引用

4th International Conference on Ubiquitous Computing and Intelligent Information Systems, ICUIS 2024

作者： Nayomi, B.Deena Divya Jyothsna, V. Department of Computer Science and Engineering Andhra Pradesh Tirupati India Department of Data Science Andhra Pradesh Tirupati India

ISBN: (纸本)9798331529635

This research study presents a thorough examination of the application of deep learning techniques in enhancing node position prediction within Vehicular Ad-Hoc Networks (VANETs). As VANETs become increasingly integral to the development of intelligent transportation systems, accurate and efficient node position prediction emerges as a critical component for ensuring optimal network functionality, safety, and traffic management. This paper explores the performance and adaptability of multiple deep learning models in the dynamic and complicated VANET environment, such as Graph Neural Networks (GNNs), Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs). We provide a detailed analysis of the existing literature, highlighting the strengths and limitations of different deep learning approaches in handling the high mobility, variable density, and recurrently changing topology of vehicular networks. Furthermore, the survey discusses the integration of contextual information, such as environmental factors and road conditions, into the prediction models, enhancing their accuracy and reliability. We also explore the implications of advanced methodologies like Transfer Learning and Federated Learning in this domain. In addition to providing insights into future research paths, the paper finishes with an assessment of the difficulties encountered in the practical application of these models, including data privacy and computing limits. The goal of this thorough investigation is to provide a roadmap for future developments in node location prediction in VANETs, making it an invaluable tool for scholars and professionals working in the subject of intelligent transportation systems. © 2024 IEEE.

关键词： Vehicular ad hoc networks

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：