检索结果-内蒙古大学图书馆

Suicide risk assessment using word-level model with dictionary-based risky posts selection

Multimedia Tools and applications 2024年第7期83卷 21435-21454页

作者： Tsai, Yun Sheng Chen, Arbee L. P. Institute of Information Systems and Applications National Tsing Hua University Hsinchu Taiwan Department of Computer Science and Information Engineering Asia University Taichung Taiwan Department of Computer Science National Tsing Hua University Hsinchu Taiwan

Suicide is a serious issue around the world and is a leading cause of death in US. In the past 20 years, the suicide rate has seen a significant increase of 35%. With the rapid development of information technology, more and more people begin to use social media to share their inner feelings. It enables social media data to be widely used for research on suicide risk assessment. However, not all social media posts are suicide related. Previous research addressed this problem with post-level attention mechanism. However, post-level attention mechanism may not find relevant suicide posts. This problem becomes more serious in the feature-based post embeddings since each post is converted into a single vector to serve as the input of the model, resulting in the loss of word-level information during training. In this paper, we addressed this problem by introducing a novel word-level model including a post-selectin layer as a solution. Firstly, we utilize a suicide keyword dictionary to identify risky posts that may be missed by the post-level attention mechanism. We then convert the words in the risky posts into word embeddings and use self-attention to generate the post embeddings for the risky posts. Finally, we pass the post embeddings to a multilayer perceptron to classify the suicide risk. We also demonstrate that the FScore used in previous studies can be reduced to a function of accuracy, which does not reflect the model performance in predicting imbalanced datasets. Therefore, we additionally adopt macro F1 score as the evaluation function. Experiment results show that our model not only outperforms previous studies in FScore performance, but also achieves macro F1 Score a nearly 4% improvement compared to previous studies. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2023.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Classification and study of music genres with multimodal Spectro-Lyrical Embeddings for Music (SLEM)

引用

Multimedia Tools and applications 2025年第7期84卷 3701-3721页

作者： Mehra, Ashman Mehra, Aryan Narang, Pratik Department of Computer Science and Information Systems Birla Institute of Technology and Science Goa Pilani India Computer Science Department Carnegie Mellon University PittsburghPA United States Department of Computer Science and Information Systems Birla Institute of Technology and Science Pilani Rajasthan Pilani India

The essence of music is inherently multi-modal – with audio and lyrics going hand in hand. However, there is very less research done to study the intricacies of the multi-modal nature of music, and its relation with genres. Our work uses this multi-modality to present spectro-lyrical embeddings for music representation (SLEM), leveraging the power of open-sourced, lightweight, and state-of-the-art deep learning vision and language models to encode songs. This work summarises extensive experimentation with over 20 deep learning-based music embeddings of a self-curated and hand-labeled multi-lingual dataset of 226 recent songs spread over 5 genres. Our aim is to study the effects of varying the weight of lyrics and spectrograms in the embeddings on the multi-class genre classification. The purpose of this study is to prove that a simple linear combination of both modalities is better than either modality alone. Our methods achieve an accuracy ranging between 81.08% to 98.60% for different genres, by using the K-nearest neighbors algorithm on the multimodal embeddings. We successfully study the intricacies of genres in this representational space, including their misclassification, visual clustering with EM-GMM, and the domain-specific meaning of the multi-modal weight for each genre with respect to ’instrumentalness’ and ’energy’ metadata. SLEM presents one of the first works on an end-to-end method that uses spectro-lyrical embeddings without hand-engineered features. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Music

来源：评论

学校读者我要写书评

暂无评论

Video captioning – a survey

引用

Multimedia Tools and applications 2025年第2期84卷 947-978页

作者： Vaishnavi, J. Narmatha, V. Department of Computer Applications SRM Institute of Science and Technology Trichy India Department of Computer and Information Science Annamalai University Annamalai Nagar Tamilnadu 608002 India

The explosion of the novel phenomenon of the combination of computer vision and Natural language processing is playing a vital role in converting the ordinary world into a more technological pool. Natural language processing and computer vision are the vanguard of Artificial intelligence with the enormous potential currently ruling more fields. The trending Accelerated research of the NLP and CV combo are cancer screening, surgical simulation, visual properties description, visual retrieval, visual description, etc. Video captioning is a task of localizing the event and various features are extracted from the particular event and then the description of the entire event is performed with Natural languages using various techniques. The task of video captioning has some evolution of various techniques from Traditional methods to Deep learning. Currently, Deep learning techniques are ruling the field. Several models are being proposed for the better enhancement of the video captioning task with the default Encoder-Decoder framework and some models with the variance of Attention mechanism and Transfer learning etc., The Attainment of every technique and the accuracy of the generated results are highly dependent on the nature of the problem, diversity, performance of chosen dataset and construction of the technique with various layers, and the amount of data split for train, validate, and test. This survey discusses the captioning types, real-time applications of video captioning, evolution of various methods in video captioning, advanced techniques, different datasets, metrics that are used to evaluate the results, and analysis of the results of the existing models for video captioning. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

Social Media-Based Surveillance systems for Health Informatics Using Machine and Deep Learning Techniques:A Comprehensive Review and Open Challenges

引用

computer Modeling in Engineering & sciences 2024年第5期139卷 1167-1202页

作者： Samina Amin Muhammad Ali Zeb Hani Alshahrani Mohammed Hamdi Mohammad Alsulami Asadullah Shaikh Institute of Computing Kohat University of Science and TechnologyKohat26000Pakistan Department of Computer Science College of Computer Science and Information SystemsNajran UniversityNajran61441Saudi Arabia Department of Information System College of Computer Science and Information SystemsNajran UniversityNajran61441Saudi Arabia

Social media(SM)based surveillance systems,combined with machine learning(ML)and deep learning(DL)techniques,have shown potential for early detection of epidemic *** review discusses the current state of SM-based surveillance methods for early epidemic outbreaks and the role of ML and DL in enhancing their ***,every year,a large amount of data related to epidemic outbreaks,particularly Twitter data is generated by *** paper outlines the theme of SM analysis for tracking health-related issues and detecting epidemic outbreaks in SM,along with the ML and DL techniques that have been configured for the detection of epidemic *** has emerged as a promising ML technique that adaptsmultiple layers of representations or features of the data and yields state-of-the-art extrapolation *** recent years,along with the success of ML and DL in many other application domains,both ML and DL are also popularly used in SM *** paper aims to provide an overview of epidemic outbreaks in SM and then outlines a comprehensive analysis of ML and DL approaches and their existing applications in SM ***,this review serves the purpose of offering suggestions,ideas,and proposals,along with highlighting the ongoing challenges in the field of early outbreak detection that still need to be addressed.

关键词： Social media epidemic machine learning deep learning health informatics pandemic

来源：评论

学校读者我要写书评

暂无评论

Employing cross-domain modelings for robust object detection in dynamic environment of autonomous vehicles

引用

Multimedia Tools and applications 2024年 1-47页

作者： Rawlley, Oshin Gupta, Shashank Kathera, Hardik Katyal, Siddharth Batwara, Yashvardhan Department of Computer Science & amp Information Systems Birla Institute of Technology and Science Rajasthan Pilani India

Object detection (OD) in Advanced Driver Assistant systems (ADAS) has been a fundamental problem especially when complex unseen cross-domain adaptations occur in real driving scenarios of autonomous Vehicles (AVs). During the sensory perception of autonomous Vehicles (AV) in the driving environment, the Deep Neural Networks (DNNs) trained on the existing large datasets fail to detect the vehicular instances in the real-world driving scenes having sophisticated dynamics. Recent advances in Generative Adversarial Networks (GAN) have been effective in generating different domain adaptations under various operational conditions of AVs, however, it lacks key-object preservation during the image-to-image translation process. Moreover, high translation discrepancy has been observed with many existing GAN frameworks when encountered with large and complex domain shifts such as night, rain, fog, etc. resulting in an increased number of false positives during vehicle detection. Motivated by the above challenges, we propose COPGAN, a cycle-object preserving cross-domain GAN framework that generates diverse variations of cross-domain mappings by translating the driving conditions of AV to a desired target domain while preserving the key objects. We fine-tune the COPGAN training with an initial step of key-feature selection so that we realize the instance-aware image translation model. It introduces a cycle-consistency loss to produce instance specific translated images in various domains. As compared to the baseline models that needed a pixel-level identification for preserving the object features, COPGAN requires instance-level annotations that are easier to acquire. We test the robustness of the object detectors SSD, Detectron, and YOLOv5 (SDY) against the synthetically-generated COPGAN images, along with AdaIN images, stylized renderings, and augmented images. The robustness of COPGAN is measured in mean performance degradation for the distorted test set (at IoU threshold =

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Focus-Shifting Attack: An Adversarial Attack That Retains Saliency Map information and Manipulates Model Explanations

引用

IEEE Transactions on Reliability 2024年第2期73卷 808-819页

作者： Huang, Qi-Xian Chiang, Lin-Kuan Chiu, Min-Yi Sun, Hung-Min National Tsing Hua University Institute of Information Systems and Applications Hsinchu30013 Taiwan National Tsing Hua University Department of Computer Science Hsinchu30013 Taiwan

With the increased use of deep learning in many fields, a question has been raised: 'How much should we trust the results generated by deep learning models?' Thus, there has been much research into the interpretations of model results, in order to open the black box of deep learning. The focus is more on interpretation than prediction in some fields such as medicine. Adversarial attacks are the most direct threats to deep learning models. They can add undetectable perturbations to the data to make the models give incorrect results, and model explanations are also susceptible to attacks. This leads to a loss of trust in explanations provided by the models, limiting the application and commercial value of deep learning. This research proposes a targeted adversarial attack algorithm that manipulates the interpretation of the model. Unlike other adversarial attacks on model interpretation, focus-shifting attack (FS Attack) can preserve the numerical depth of the original saliency map without specifying a perturbation budget. Experiments have shown that the FS Attack has a higher degree of image similarity and misleading interpretation than other adversarial attacks, and the property of preserving the numerical depth of the original saliency map makes it more difficult to detect. This study uses several common explanation methods as experimental subjects to investigate how these explanations can be manipulated and evaluate the effectiveness of the attack under different conditions. Under a particular interpretation, the FS Attack has a highly successful attack rate of 94.6%, which is a critical adversarial attack. © 1963-2012 IEEE.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

Parkinson’s disease Severity Detection Based on Fused features with tuned Residual Deep Convolutional Neural Network

引用

Multimedia Tools and applications 2025年第14期84卷 13111-13131页

作者： Vishwakarma, Gagan Nandanwar, Amit Kumar Thakur, Ghanshyam Singh Computer Science & Engineering Indian Institute of Information Technology MP Bhopal India Computer Science & Engineering Maulana Azad National Institute of Technology MP Bhopal India Department of Computer Applications Maulana Azad National Institute of Technology MP Bhopal India

Parkinson's disease (PD) diagnosis involves the assessment of a variety of motor and non-motor symptoms. To accurately diagnose PD, it is necessary to differentiate its symptoms from those of other conditions. During the diagnosis process, clinicians must consider gait abnormalities as one of the key symptoms. However, relying solely on clinical judgment for the diagnosis of PD may have limitations. In this work, we have proposed a Multikernel SVM (MK-SVM) based Parkinson's severity analysis method. Initially, an adaptive median filter is used to reduce the noise in the gait image. Then, to extract the features, we propose a dual-layer method that employs stacked autoencoders (SAEs) for local feature extraction and residual deep convolution neural networks (RDCNN) for global feature extraction. The hyper-parameters from the RDCNN are optimized using the Enhanced walrus optimization algorithm (EWO). In the final step, we use MK-SVM to classify PD according to severity. This model is evaluated using the GAIT-IT and GAIT-IST datasets. With the help of Python, we evaluated the proposed model in terms of precision, accuracy, recall, specificity, and F-scores, and the model achieved a maximum accuracy of 97.3%. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Early Detection of Colletotrichum Kahawae Disease in Coffee Cherry Based on computer Vision Techniques

引用

computer Modeling in Engineering & sciences 2024年第4期139卷 759-782页

作者： Raveena Selvanarayanan Surendran Rajendran Youseef Alotaibi Department of Computer Science and Engineering Saveetha School of EngineeringSaveetha Institute of Medical and Technical ScienceChennai602105India Department of Computer Science College of Computer and Information SystemsUmm Al-Qura UniversityMakkah21955Saudi Arabia

Colletotrichum kahawae(Coffee Berry Disease)spreads through spores that can be carried by wind,rain,and insects affecting coffee plantations,and causes 80%yield losses and poor-quality coffee *** deadly disease is hard to control because wind,rain,and insects carry *** researchers utilized a deep learning system to identify CBD in coffee cherries at three growth stages and classify photographs of infected and uninfected cherries with 93%accuracy using a random forest *** the dataset is too small and noisy,the algorithm may not learn data patterns and generate accurate *** overcome the existing challenge,early detection of Colletotrichum Kahawae disease in coffee cherries requires automated processes,prompt recognition,and accurate *** proposed methodology selects CBD image datasets through four different stages for training and *** to train a model on datasets of coffee berries,with each image labeled as healthy or *** themodel is trained,SHAP algorithmto figure out which features were essential formaking predictions with the proposed *** of these characteristics were the cherry’s colour,whether it had spots or other damage,and how big the Lesions *** inception is important for classification to virtualize the relationship between the colour of the berry is correlated with the presence of *** evaluate themodel’s performance andmitigate excess fitting,a 10-fold cross-validation approach is *** involves partitioning the dataset into ten subsets,training the model on each subset,and evaluating its *** comparison to other contemporary methodologies,the model put forth achieved an accuracy of 98.56%.

关键词： computer vision coffee berry disease colletotrichum kahawae XG boost shapley additive explanations

来源：评论

学校读者我要写书评

暂无评论

Thrips Detection and Monitoring System in Mango Greenhouses Based on Deep Learning Image Recognition

引用

IEEE Transactions on Instrumentation and Measurement 2024年 73卷 1-18页

作者： Wu, Ge-Han Huang, Nen-Fu Huang, Yu-Hsiang Chan, Pin-Cheng Bhat, Showkat Ahmad National Tsing Hua University Department Of Computer Science Hsinchu1300044 Taiwan National Tsing Hua University Institute Of Information Systems And Applications Hsinchu1300044 Taiwan National Tsing Hua University Center Of Innovative Incubator Department Of Computer Science Hsinchu1300044 Taiwan

The ability of thrips to infect plants with viruses, oviposition, and feeding makes it one of the most damaging agricultural pests in the world. Due to their invasiveness, many pest species pose serious challenges to growers of a wide variety of crops especially Mango in both open fields and greenhouses globally. For detecting various pests, sticky paper traps have been widely used. It is essential to inspect sticky paper traps as part of an integrated pest management (IPM) process. Identification and counting the pests caught in the traps is a laborious process. A timely and effective strategy is therefore required to resolve the issue. This study proposes and develops a complete and feasible integrated pest detection and monitoring system (IPDMS) by combining an ultrahigh resolution pest monitoring device coupled with a deep learning-based object detection model. The developed system was able to collect sticky paper images with a resolution of 64 megapixels and an AWS database for analysis. Test results from two Mango greenhouses show that the model can detect thrips with 90% accuracy and an F1 -score of 0.95. Additionally, a web-based user interface was developed to provide visual image recognition results and thrip population analysis. Thus, the proposed IPDMS addresses the significant problems with pest detection systems (PDSs) by providing real-time information on pest occurrences in greenhouses. Thus, it makes pest detection and monitoring in agriculture efficient. © 1963-2012 IEEE.

关键词： Image recognition

来源：评论

学校读者我要写书评

暂无评论

Utilizing Multi-Objective Evolutionary Algorithms to Optimize Open Source Software Release Management

引用

IEEE Access 2023年 11卷 112248-112262页

作者： Yu, Chen-Yang Huang, Chin-Yu Institute of Information Systems and Applications National Tsing Hua University Hsinchu300044 Taiwan National Tsing Hua University Department of Computer Science Hsinchu300044 Taiwan

Scheduling and resource allocation for Open Source Software (OSS) product development pose crucial and challenging tasks due to program size and resource limitations. The properties of OSS further complicate product assessment and maintenance for developers. This paper proposes a model for the iterative and multi-release OSS development process. Unlike traditional methods that oversimplify the problem by reducing the multi-decision space into a single-objective optimization problem, our approach suggests employing Multi-objective Evolutionary Algorithms (MOEAs) to solve the Optimal Release Time Planning Problem, enabling the simultaneous maximization of reliability and minimization of cost. We consider testing cost and system reliability, two critical dimensions, as the primary objectives, while also incorporating testing resource consumption as the third objective. The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is chosen as the primary method for its effectiveness in MOEAs, with a special scenario outlined in the paper where NSGA-II may not guarantee optimal solutions. Demonstrating the practicality of our proposed method, we utilize open source data to assess release time and illustrate its superiority to NSGA-II. Numerical examples further showcase the model's effectiveness. © 2013 IEEE.

关键词： Multiobjective optimization

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：