检索结果-内蒙古大学图书馆

Robust video question answering via contrastive cross-modality representation learning

science China(Information sciences) 2024年第10期67卷 211-226页

作者： Xun YANG Jianming ZENG Dan GUO Shanshan WANG Jianfeng DONG Meng WANG School of Information Science and Technology University of Science and Technology of China Institute of Artificial Intelligence Hefei Comprehensive National Science Center School of Computer Science and Information Engineering Hefei University of Technology Institutes of Physical Science and Information Technology Anhui University School of Computer Science and Technology Zhejiang Gongshang University

Video question answering(VideoQA) is a challenging yet important task that requires a joint understanding of low-level video content and high-level textual semantics. Despite the promising progress of existing efforts, recent studies revealed that current VideoQA models mostly tend to over-rely on the superficial correlations rooted in the dataset bias while overlooking the key video content, thus leading to unreliable results. Effectively understanding and modeling the temporal and semantic characteristics of a given video for robust VideoQA is crucial but, to our knowledge, has not been well investigated. To fill the research gap, we propose a robust VideoQA framework that can effectively model the cross-modality fusion and enforce the model to focus on the temporal and global content of videos when making a QA decision instead of exploiting the shortcuts in datasets. Specifically, we design a self-supervised contrastive learning objective to contrast the positive and negative pairs of multimodal input, where the fused representation of the original multimodal input is enforced to be closer to that of the intervened input based on video perturbation. We expect the fused representation to focus more on the global context of videos rather than some static keyframes. Moreover, we introduce an effective temporal order regularization to enforce the inherent sequential structure of videos for video representation. We also design a Kullback-Leibler divergence-based perturbation invariance regularization of the predicted answer distribution to improve the robustness of the model against temporal content perturbation of videos. Our method is model-agnostic and can be easily compatible with various VideoQA backbones. Extensive experimental results and analyses on several public datasets show the advantage of our method over the state-of-the-art methods in terms of both accuracy and robustness.

关键词： video question answering cross-modality fusion contrastive learning cross-media reasoning

来源：评论

学校读者我要写书评

暂无评论

PotatoLeafGuard: A Novel Hybrid CNN Architecture for Advanced Detection of Potato Leaf Diseases

PotatoLeafGuard: A Novel Hybrid CNN Architecture for Advance...

引用

2024 International Conference on Emerging Innovations and Advanced Computing, INNOCOMP 2024

作者： Sharma, Meenakshi Raman Babal Sharma, Naman Varun Navpreet Global Group of Institutes Department of Computer Science Engineering Amritsar India

ISBN: (纸本)9798350376470

Potatoes assume a crucial role as a global food crop, contending with challenges posed by various pathogens affecting their growth. Ranked as the third-largest agricultural food crop, potatoes are indispensable for delivering essential nutrients. One significant threat to potato health arises from leaf diseases. The accurate detection of these diseases is imperative for effective intervention. This research introduces a novel hybrid model integrating deep learning and transfer learning to predict the risks associated with potato leaf diseases. To train the model, a dataset comprising 3,076 images categorized into seven classes is used. These classes delineate various types of leaf diseases, including those affected by viruses, bacteria, fungi, pests, nematodes, phytophthora, and healthy leaves. The methodology leverages Convolutional Neural Networks (CNNs), recognized for their efficacy in identifying potato leaf diseases through image analysis. To enhance accuracy, Transfer Learning algorithms such as VGG16 and Xception are incorporated. This hybrid model addresses the intricate classification challenges arising from the distinct characteristics and data limitations inherent in potato leaf diseases. Furthermore, data augmentation techniques are employed in conjunction with CNNs to mitigate classification challenges associated with unique features and limited data. The stacking of outputs from these models substantially enhances the accuracy of potato leaf disease detection. In summary, this research endeavors to provide a novel methodology for the precise classification of diverse types of potato leaf diseases, contributing to improved crop management and food security. The proposed model achieves an overall classification accuracy of 94.70%, accompanied by commendable precision, recall, and F1-score values of 93.75%, 93.6%, and 92.75%, respectively. These results validate the model's efficacy in advancing research on potato leaf disease detection, attesting to its

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

A Comparative Analysis of CNN and Transfer Learning Algorithms in Predicting Alzheimer's Disease Stages Using Neuroimaging

A Comparative Analysis of CNN and Transfer Learning Algorith...

引用

2024 OPJU International Technology Conference on Smart Computing for Innovation and Advancement in Industry 4.0, OTCON 2024

作者： Sharma, Meenakshi Bhagat, Bhavneet Kumari, Vandana Bhumi Sharma, Naman Singh, Dipti Global Group of Institutes Department of Computer Science Engineering Amritsar India Guru Nanak Dev University Department of Computer Science Engineering Amritsar India

ISBN: (纸本)9798350373783

Alzheimer's disease (AD) prediction plays a critical role in early detection and intervention, allowing for better patient care and management. This paper investigates the predictive capabilities and state-of-art of Convolutional Neural Networks (CNN) and transfer learning models, including Inception V3, Resnet-18, and Resnet-50, for the progression of Alzheimer's disease (AD). The Alzheimer's disease Neuroimaging Initiative (ADNI) dataset used in this study, which includes multimodal neuroimaging, clinical, genetic and demographic data, the research aims to provide accurate predictions at various stages of the EA. This dataset enables robust training and evaluation of deep learning algorithms to predict AD progression. The data undergoes preprocessing and feature engineering to prepare a robust data set. Evaluations covering accuracy, precision, recall and F1 score, using k-fold cross-validation, are performed to measure the performance of the chosen algorithms. CNN and Inception V3 demonstrate remarkable performance with an accuracy of 93.75% and 94%. Comparative analyzes include exploration of Inception V3, Resnet-18, and Resnet-50, shedding light on their individual contributions to AD prediction. This paper contributes to the development of reliable and efficient deep learning approaches for early prediction and diagnosis of AD, with the potential to improve patient outcomes. Comparing the predictive capabilities of CNNs and transfer learning models provides valuable insights into their effectiveness in modeling AD progression. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Brain Tumor Detection and Localization from MRI Scan Images 14

Brain Tumor Detection and Localization from MRI Scan Images

引用

14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023

作者： Sharma, Richa Kumari, Radhika Thakur, Bhanu Kumar, Kuldeep Chandigarh University Computer Science and Engineering Chandigarh India Global Group of Institutes Computer Science and Engineering Amritsar India

ISBN: (纸本)9798350335095

A tumor is a tissue mass produced by an accumulation of uncontrolled cells that keeps growing. A brain cancer is a tumor that is either framed in the cerebrum or has spread there. In this research, a model is proposed that can accurately segments pictures using the PSO methods, extracts features using the PCA strategies, and classifies tumors using the CNN. A mind cancer is a growth which is either framed in the cerebrum or has relocated No essential driver has been recognized for the arrangement of growths in the cerebrum till date. However, cancers in the mind are not extremely normal (Worldwide cerebrum growths make up just 1.8% of complete detailed cancers), the death pace of threatening cerebrum cancers is exceptionally high because of the way that the growth development is in the most basic organ of the body. Subsequently, it is of most extreme significance to precisely recognize mind cancers at beginning phases to bring down the death rate. We have accordingly proposed a space variation based Deep Neural Network which will evaluate cerebrum growths from MRI checks for the administration of mind cancer conclusion. In this review, we have executed a model that identify the pictures by utilizing ResNet based profound brain network which has previously been prepared on huge picture datasets and move this preparing information into our own information. To augment the presentation, I have additionally tweaked it in light of our concern set. © 2023 IEEE.

关键词： Tumors

来源：评论

学校读者我要写书评

暂无评论

Sentimental Analysis of Stock Market via Twitter

Sentimental Analysis of Stock Market via Twitter

引用

International Conference on Machine Learning and Big Data Analytics, ICMLBDA 2022

作者： Lakshmi, S.V.S.S. Vineetha Sai Durga, T. Tejaswi, N.V.L. Yeshwanth Sai Kumar, A. Department of Computer Science and Engineering Anil Neerukonda Institutes of Technology and Sciences Visakhapatnam India

ISBN: (纸本)9783031151743

The immense increase in technology has led to unbelievable changes which include, mainly, global digitization connecting markets around the world on a global scale. People all over the world are now interested in investing their money in the global market via the stock market. The stock market is a place where everyone can invest their money into shares of publicly held companies. While the stock market is such a buzz word, very few people actually know how and when to invest and when to buy and sell shares. The stock market runs on predictions, precisely made predictions, but it changes so constantly either via word of mouth or insider knowledge;therefore, it is necessary to be informed about the market at all times in order to make profitable investments (Saurabh Singh, Twitter sentimental analysis using machine learning. IJSRCSEIT, 2020). This paper proposes a machine learning approach to overcome all the issues with investing in the stock market. It proposes a sentimental analysis that is made on tweets that are collected from Twitter in real time, in order to make a precise prediction on whether a particular stock investment is positive, negative, or neutral for that particular [1]. In addition to this, the paper also proposes a tool with which the sentiments of a stock can be checked, as well as containing important information about the stock, the daily reports of the market mood, the specific details on the stock, and so on. The analysis can be done by using the Naïve Bayes approach using python and includes web scraping, matplotlib library, and other python libraries necessary. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Financial markets

来源：评论

学校读者我要写书评

暂无评论

GOMAA-Geo: GOal Modality Agnostic Active Geo-localization 38

GOMAA-Geo: GOal Modality Agnostic Active Geo-localization

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Sarkar, Anindya Sastry, Srikumar Pirinen, Aleksis Zhang, Chongjie Jacobs, Nathan Vorobeychik, Yevgeniy Department of Computer Science and Engineering Washington University in St. Louis United States RISE Research Institutes of Sweden Sweden Sweden

We consider the task of active geo-localization (AGL) in which an agent uses a sequence of visual cues observed during aerial navigation to find a target specified through multiple possible modalities. This could emulate a UAV involved in a search-and-rescue operation navigating through an area, observing a stream of aerial images as it goes. The AGL task is associated with two important challenges. Firstly, an agent must deal with a goal specification in one of multiple modalities (e.g., through a natural language description) while the search cues are provided in other modalities (aerial imagery). The second challenge is limited localization time (e.g., limited battery life, urgency) so that the goal must be localized as efficiently as possible, i.e. the agent must effectively leverage its sequentially observed aerial views when searching for the goal. To address these challenges, we propose GOMAA-Geo - a goal modality agnostic active geo-localization agent - for zero-shot generalization between different goal modalities. Our approach combines cross-modality contrastive learning to align representations across modalities with supervised foundation model pretraining and reinforcement learning to obtain highly effective navigation and localization policies. Through extensive evaluations, we show that GOMAA-Geo outperforms alternative learnable approaches and that it generalizes across datasets - e.g., to disaster-hit areas without seeing a single disaster scenario during training - and goal modalities - e.g., to ground-level imagery or textual descriptions, despite only being trained with goals specified as aerial views. Our code is available at: https://***/mvrl/GOMAA-Geo. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

MULTISOURCE DUAL UNSUPERVISED DOMAIN ADAPTATION METHOD FOR REMAINING LIFE PREDICTION UNDER MULTIPLE OPERATING CONDITIONS 14

MULTISOURCE DUAL UNSUPERVISED DOMAIN ADAPTATION METHOD FOR R...

引用

14th International Conference on Quality, Reliability, Risk, Maintenance, and Safety engineering, QR2MSE 2024

作者： Zhou, Ningxu Lyu, Yi Shen, Zaichen Xiao, Xu Fan, Ruhui School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China School of Computer University of Electronic Science and Technology of China Zhongshan Institutes Zhongshan China School of Automation Guangdong University of Technology Guangzhou China

ISBN: (纸本)9781837242672

Aero engines, with their complex structures and intricate gas paths, are a crucial component of the aviation system. Traditional machine learning-based methods for predicting the remaining useful life (RUL) of rolling bearings often struggle to provide adequate feature representation and adaptive extraction due to complex operating conditions and challenges in obtaining degradation data. Given that traditional machine learning methods require substantial amounts of labeled data, they frequently fail to yield satisfactory RUL prediction results. To overcome these limitations, domain adaptation methods have been employed, facilitating the transfer of knowledge from a well-labeled source domain to an unlabeled target domain. Additionally, to augment the available data, a multisource domain transfer approach is proposed, which transfers data from multiple source domains to the target domain intended for prediction. To enhance the effectiveness of this multisource domain transfer method, a dual alignment strategy is introduced for aligning features at both the feature and semantic levels, thereby improving predictive performance. The model incorporates a convolutional neural network (CNN) and long short-term memory (LSTM) as its backbone, leveraging CNN's robust feature extraction capabilities combined with LSTM’s proficiency in processing time-dependent data. Experiments conducted on the CMAPSS dataset, along with comparisons and ablation studies against current popular multisource domain adaptation methods, demonstrate the superiority of the proposed approach. © The Institution of engineering & Technology 2024.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Leveraging hierarchical semantic‐emotional memory in emotional conversation generation

引用

CAAI Transactions on Intelligence Technology 2023年第3期8卷 824-835页

作者： Min Yang Zhenwei Wang Qiancheng Xu Chengming Li Ruifeng Xu Shenzhen Institutes of Advanced Technology Chinese Academy of SciencesShenzhenChina School of Electrical and Computer Engineering Georgia Institute of TechnologyAtlantaUSA School of Intelligent Systems Engineering Sun Yat‐Sen UniversityGuangzhouChina School of Computer Science Harbin Institute of Technology(Shenzhen)ShenzhenChina

Handling emotions in human‐computer dialogues has emerged as a challenging task which requires artificial intelligence systems to generate emotional responses by jointly perceiving the emotion involved in the input posts and incorporating it into the gener-ation of semantically coherent and emotionally reasonable ***,most previous works generate emotional responses solely from input posts,which do not take full advantage of the training corpus and suffer from generating generic *** this study,we introduce a hierarchical semantic‐emotional memory module for emotional conversation generation(called HSEMEC),which can learn abstract semantic conver-sation patterns and emotional information from the large training *** learnt semantic and emotional knowledge helps to enrich the post representation and assist the emotional conversation *** experiments on a large real‐world conversation corpus show that HSEMEC can outperform the strong baselines on both automatic and manual *** reproducibility,we release the code and data publicly at:https://***/siat‐nlp/HSEMEC‐code‐data.

关键词： deep learning emotional conversation generation semantic‐emotional memory

来源：评论

学校读者我要写书评

暂无评论

ICA-Net:improving class activation for weakly supervised semantic segmentation via joint contrastive and simulation learning

引用

Optoelectronics Letters 2025年第3期21卷 188-192页

作者： YE Zhuang LIU Ruyu SUN Bo School of Information Engineering Chengdu Aviation Vocational and Technical CollegeChengdu610100China School of Computer Science and Engineering Hangzhou Normal UniversityHangzhou311121China Quanzhou Institute of Equipment Manufacturing Haixi InstitutesChinese Academy of SciencesQuanzhou362200China

In the field of optoelectronics,certain types of data may be difficult to accurately annotate,such as high-resolution optoelectronic imaging or imaging in certain special spectral *** supervised learning can provide a more reliable approach in these *** popular approaches mainly adopt the classification-based class activation maps(CAM)as initial pseudo labels to solve the task.

关键词： high resolution imaging supervised learning class activation maps joint contrastive simulation learning special spectral ranges weakly supervised learning optoelectronics

来源：评论

学校读者我要写书评

暂无评论

Enhanced Electrocardiogram Data Compression and Classification Using 2D Convolutional Neural Networks 5

Enhanced Electrocardiogram Data Compression and Classificati...

引用

5th International Conference on Emerging Systems and Intelligent Computing, ESIC 2025

作者： Yadav, Amrendra Singh Mehta, Preeti Mishra, Deepti Barik, Rabindra Kumar AVM-Indian Institute of Information Technology and Management Department of Computer Science and Engineering Gwalior India National Institutes of Technology Delhi Department of Computer Science and Engineering Delhi India Kiit Deemed to be University School of Economics and Commerce Bhubaneshwar India Kiit Deemed to be University School of Computer Application Bhubaneshwar India

ISBN: (纸本)9798331522100

The task of electrocardiogram (ECG) classification is a combination of data compression and classification of signals based on the patient's health. This paper introduces a novel technique for compressing electrocardiogram (ECG) images and utilizing these compressed images for accurate classification. We transform ECG signals into images and compress them to reduce their size while preserving classification accuracy significantly. Our approach employs a 2-dimensional Convolutional Neural Network (2-D CNN) model and achieves an impressive accuracy of 97% even after compressing the images by up to 76%. The proposed technique was tested on the widely-used MIT-BIH Arrhythmia Dataset, demonstrating its robustness and efficacy. Maintaining high classification accuracy despite significant image compression is crucial for efficient storage and transmission in resource-constrained environments. This method holds substantial potential for real-time monitoring and diagnosis in telemedicine applications, emphasizing the importance of our contribution to the field. © 2025 IEEE.

关键词： Electrocardiography

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：