检索结果-内蒙古大学图书馆

STPM: Spatial-Temporal Token Pruning and Merging for Complex Activity Recognition

IEEE Transactions on Circuits and Systems for Video technology 2024年第6期35卷 5407-5418页

作者： Su, Yumeng Zhang, Jiachao Yan, Rui Li, Pengpeng Xie, Guosen Shu, Xiangbo Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing210094 China Nanjing Institute of Technology Artificial Intelligence Industrial Technology Research Institute Nanjing211167 China Nanjing University Department of Computer Science and Technology Nanjing210023 China

Lightweight video representation techniques have advanced significantly for simple activity recognition, but they still encounter several issues when applied to complex activity recognition: (i) The presence of numerous individuals and varying spatial positions makes it difficult for traditional token pruning methods to maintain accuracy. (ii) Simply discarding entire frames may result in the loss of crucial clues. (iii) To maintain parallel computing, applying the same pruning rate to every frame leads to significant redundancy in frames with low information content. To this end, we propose a lightweight and novel Spatial-Temporal Token Pruning and Merging (STPM) framework, specifically designed for complex action videos where human actors occupy a small spatial resolution within video frames. Our framework considers two critical factors: semantic importance and spatial-temporal redundancy, to further reduce overhead. For semantic importance, STPM captures class-specific attention scores by learning multiple class tokens within the transformer to guide token pruning. For spatial-temporal redundancy, STPM employs an anchor graph and temporal attention to perform spatial and temporal token merging, preserving appearance and temporal cues while eliminating semantic duplication and redundancy. We conduct extensive experiments on JRDB-PAR primarily using recently introduced video transformer backbones, e.g., MViT and ViT. Our framework achieves similar results while requiring 40% less computation. © 1991-2012 IEEE.

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

A Smart Contract-Based Decentralised Crowdfunding Platform for Increased Security and Transparency 1

A Smart Contract-Based Decentralised Crowdfunding Platform f...

引用

1st International Conference on Cognitive, Green and Ubiquitous Computing, IC-CGU 2024

作者： Antad, Sonali Vidhale, Arpit Vikhar, Shivani Thite, Sanket Warke, Aditya Vishwakarma Institute of Technology Department of Computer Engineering Pune India

ISBN: (纸本)9798350361780

In the contemporary landscape of financial systems and digital innovations, the convergence of blockchain technology and crowdfunding models has unveiled a promising avenue for revolutionizing traditional fundraising platforms. This research paper delves into the intricate fabric of designing and implementing a robust crowdfunding platform underpinned by the decentralized principles of blockchain technology integrated with Artificial Intelligence. The research embarks on an in-depth exploration of the underlying mechanics of blockchain, elucidating its distributed consensus, immutability, and transparency features. A comprehensive survey of existing literature pertaining to crowdfunding and blockchain integration is presented, establishing a solid foundation for the proposed platform's development. The conceptual framework delineates the architectural blueprint of the decentralized crowdfunding platform, intricately woven with smart contracts, artificial intelligence, cryptographic protocols, and a personalized recommendation system. The amalgamation of these components engenders an environment that ensures trust, security, and automation throughout the crowdfunding lifecycle. Moreover, this paper delves into the pivotal aspect of tokenization within the blockchain-based crowdfunding ecosystem. By implementing these cases against theoretical frameworks, the paper elucidates a comprehensive set of best practices and potential pitfalls that can guide the development and deployment of a successful blockchain-based crowdfunding platform. © 2024 IEEE.

关键词： Smart contract

来源：评论

学校读者我要写书评

暂无评论

Harnessing Natural Entity Recognition for Intuitive Spreadsheet Control: A Voice-Driven Google Sheets Extension 5

Harnessing Natural Entity Recognition for Intuitive Spreadsh...

引用

5th IEEE International Conference for Emerging technology, INCET 2024

作者： Joglekar, Pushkar Jha, Divyanshu Dhorage, Prathamesh Thakkar, Dhruv Dhumal, Om Department of Computer Engineering Vishwakarma Institute of Technology Pune India

ISBN: (纸本)9798350361155

This paper introduces an innovative Voice-Controlled Google Sheets Extension employing Natural Language Processing (NLP) capabilities. The system seamlessly integrates a Chrome extension and a meticulously trained Named Entity Recognition (NER) model. Trained on custom datasets, the NER model proficiently extracts crucial entities, such as command types, cell references, and values from user speech, enabling accurate interpretation of complex commands. The extension enhances user interaction with spreadsheets, offering a voice-driven interface that simplifies spreadsheet management. The backend, utilizing Flask-based REST APIs, facilitates seamless communication with the Google Sheets API, ensuring precise execution of actions dictated by the interpreted commands. Experimental results underscore the system's accuracy and efficiency, establishing its potential as an intelligent productivity tool. This research contributes to the evolution of smart computing interfaces, providing users with an intuitive and accessible means of interacting with spreadsheet applications. © 2024 IEEE.

关键词： Natural language processing systems

来源：评论

学校读者我要写书评

暂无评论

Generative Adversarial Networks (GANs) for Education: State-Of-Art and Applications

Generative Adversarial Networks (GANs) for Education: State-...

引用

International Conference on Power engineering and Intelligent Systems, PEIS 2024

作者： Kavyashree, K.N. Shidaganti, Ganeshayya Department of Computer Science and Engineering Ramaiah Institute of Technology Bangalore India

ISBN: (纸本)9789819767090

In the ever-changing educational landscape, technology integration has become essential to improving students’ understanding of basic topics. The COVID-19 pandemic hastened the transition to online education while posing new difficulties for teachers and students alike. The disruption of traditional classroom dynamics was exacerbated by the lack of digital innovation and interactive instructional tools, which made student participation less effective. Traditional teaching paradigms must make way for ones that encourage self-directed learning and fortify technical infrastructure in order to handle these difficulties. Generative Adversarial Networks (GANs), a cutting-edge use of deep learning, are starting to revolutionize educational technology. With the use of GANs, a range of features that are tailored to the individual learning preferences of each student may be created, thereby bridging knowledge gaps and providing a more dynamic and interesting learning environment for students. This research investigates several state-of-the-art features that leverage technology linked to GANs to enhance students’ virtual learning environments. The novel applications of generative adversarial networks are explored, with a focus on applications for student observation, academic performance, questioning and answering, and personalized learning. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

HealthVault: A Blockchain-based Patient Health Records Management System 9

HealthVault: A Blockchain-based Patient Health Records Manag...

引用

9th IEEE International Conference for Convergence in technology, I2CT 2024

作者： Jaybhaye, Sangita Desai, Rutuja Choudhari, Gauri Dashetwar, Isha Dagale, Pratik Vishwakarma Institute of Technology Department of Computer Engineering Pune India

ISBN: (纸本)9798350394474

This paper describes a blockchain-based patient record management system that combines a React online user interface with a MongoDB database, IPFS, Hardhat, MetaMask, IPFS smart contracts, and the Ethereum blockchain. Existing systems are often subject to issues such as single-point failures, data loss, high costs and limited scalability. Failure of a server or data center can lead to data loss and system downtime. The proposed work intends to address problems with existing patient records management systems, including data security, privacy, and accessibility. The system uses Ethereum decentralized blockchain with smart contracts to store and manage patient records ensuring data privacy, security and immutability. Ether is the crypto-currency used to pay transaction fees. MetaMask is used to facilitate secure transactions between users and the blockchain network whereas IPFS is used to store large files such as medical images, videos and documents. The web interface provides an intuitive and user-friendly platform for healthcare providers and patients to interact with the system and manage their records. The proposed system provides a secure and efficient way to manage patient records enabling healthcare providers to easily access patient records while maintaining the privacy and security of sensitive medical information. © 2024 IEEE.

关键词： Smart contract

来源：评论

学校读者我要写书评

暂无评论

Simulating Application Driven Mobile D2D underlay Networks using Gym API 17

Simulating Application Driven Mobile D2D underlay Networks u...

引用

17th International Conference on COMmunication Systems and NETworkS, COMSNETS 2025

作者： Batabyal, Suvadip National Institute of Technology Department of Computer Science & Engineering Durgapur India

ISBN: (纸本)9798331531195

Device-to-device (D2D) communication underlay cellular network improves spectral efficiency by reusing cellular resources over short and low-power links between the devices. For this, the D2D paradigm is considered to be an integral part of all kinds of future networks and is an active area of research at different levels. This requires a simple, configurable, scalable, and realistic evaluation of the protocols proposed for the D2D networking paradigm. Although there exists a few platform to evaluate a D2D underlay network, they have one or more of the following drawbacks: (i) long development time and complex environment, (ii) limited support for mobility, (iii) no support for application specific development and evaluation. In this work, we developed a framework for simulating an application-driven mobile D2D underlay network using the Gym API, where D2D connections are made spontaneously based on the application requirements. The proposed framework includes modules for generating application-specific traffic, buffer for transmission and caching of application data, and mobility management. This complements several earlier works on application (or utility) driven resource allocation in D2D underlay network. The framework is used to evaluate a D2D network using different mobility models and performance of cache based video streaming in a D2D underlay network. Finally, the framework is compared with ns-3 and OMNET++ simulators and is observed to have lower CPU utilization and run-time under the same network settings. © 2025 IEEE.

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

Enhancing Used Car Price Predictions with Machine Learning-Based Damage Detection 5

Enhancing Used Car Price Predictions with Machine Learning-B...

引用

5th IEEE Global Conference for Advancement in technology, GCAT 2024

作者： Kulkarni, Radhika V. Thopate, Kaushalya Khatib, Fatima Dixit, Amogh Ingle, Ayush Kanathia, Arnav Vishwakarma Institute of Technology Department of Computer Engineering Pune India

ISBN: (纸本)9798350376685

Due to the rising sale of cars, the price of secondhand cars has evolved into a significant concern, which may negatively impact sustainability. Using online platforms to determine the price of used vehicles has become increasingly popular to ease the gap between buyers and sellers. This study contributes to the improvement of the used car industry by enhancing the accuracy of damage detection and price prediction and building trust among buyers and sellers by incorporating a machine learning-based system. It aims to automate the manual work required while buying and selling cars, including damage assessment and price adjustment based on features of the car provided by the user. This paper demonstrates a system that accepts user input in the form of text and images and employs You Only Look Once (YOLO) to detect damages in uploaded images of used cars. The cost of these vehicles is predicted based on relevant features by applying three regression and boosting algorithms to publicly available datasets and selecting the best one. © 2024 IEEE.

关键词： car damage detection car price prediction object detection regression YOLO

来源：评论

学校读者我要写书评

暂无评论

Online News: Event Detection and Community Identification 8th

Online News: Event Detection and Community Identification

引用

8th International Conference on Information and Communication technology for Intelligent Systems, ICTIS 2024

作者： Kulkarni, Radhika V. Suryawanshi, Snehal Shirsath, Aditya Sonone, Swati Phapale, Siddhi Department of Computer Engineering Vishwakarma Institute of Technology Pune India

ISBN: (纸本)9789819766833

Nowadays, reading online news is very common among people. The amount of information on the web is growing at a rapid pace, which is driving up the consumption of online news. As a result, a system to gather and disseminate significant real-world events from massive amounts of web data is now required. In addition, a lot of businesses rely their judgements on global events that happen every day. One of the most important tasks is event recognition, which is finding news articles about the same occurrence and grouping them together. This method improves user navigation and gives news pieces a conceptual structure. To group linked news documents, it makes use of TF-IDF analysis, time-series analysis, and community detection. For interactive investigation and analysis, the system interfaces with an easy-to-use website that offers sentiment analysis and 3D visualizations for improved interpretability. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Community detection Event detection News stream

来源：评论

学校读者我要写书评

暂无评论

Vitunix: A Lightweight and Secure Linux Distribution 1

引用

International Conference on Recent Developments in Cyber Security, ReDCySec 2023

作者： Shinde, Sandip Suram, Gourav Khadde, Mayur Gade, Shrutika Arthamwar, Vaishnavi Pardeshi, Palak Department of Computer Engineering Vishwakarma Institute of Technology Pune India

ISBN: (数字)9789819998111

ISBN: (纸本)9789819998104

Vitunix is a lightweight Linux distribution designed to be fast, secure, and highly customizable. Unlike traditional Linux distributions that are often bloated and resource intensive, Vitunix is optimized for speed and can run on older hardware without sacrificing performance. The distribution uses a combination of bspwm and openbox window managers to provide users with a high level of customizability. Security is also a priority for Vitunix. The distribution uses Polkit to manage system-level access controls, allowing users to specify fine-grained access controls that limit the actions that specific users or groups can perform on the system. In this paper, we introduced Vitunix, including its architecture, installation process, and package management system. We also compare Vitunix to other lightweight Linux distributions and highlight its unique strengths and weaknesses. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Open source software

来源：评论

学校读者我要写书评

暂无评论

Real-Time Traffic Management Using Deep Learning and Object Detection using YOLOv8 15

Real-Time Traffic Management Using Deep Learning and Object ...

引用

15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024

作者： Phatangare, Sheetal Sakpal, Rahul A. Kasurde, Soham N. Punde, Shubhan S. Khan, Saif S. Vishwakarma Institute of Technology Department of Computer Engineering Pune India

ISBN: (纸本)9798350370249

As urban traffic management challenges continue to grow, there is an urgent demand for advanced technologies to improve real-time monitoring and control systems. This study introduces an innovative method for realtime traffic management by combining deep learning with YOLOv8-based object detection. By leveraging the speed and accuracy of YOLOv8, our system achieves precise identification of vehicles and pedestrians in dynamic urban settings. The model seamlessly integrates with existing surveillance infrastructure, allowing for intelligent analysis of live video streams to monitor traffic flow and congestion. The system's adaptability is showcased through training on diverse datasets, and its impact extends to optimizing traffic signal control for enhanced transportation efficiency. Results suggest the potential of our approach to transform urban traffic control systems with improved real-time decision-making capabilities. © 2024 IEEE.

关键词： computer Vision Intelligent transport systems Machine Learning Object detection Real-time traffic management Traffic flow analysis Traffic management YOLOv8

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：