检索结果-内蒙古大学图书馆

2024 conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Liao, Yuan-Hong Mahmood, Rafid Fidler, Sanja Acuna, David University of Toronto Vector Institute 2 NVIDIA University of Ottawa Canada

ISBN: (纸本)9798891761643

Despite recent advances demonstrating vision-language models' (vLMs) abilities to describe complex relationships in images using natural language, their capability to quantitatively reason about object sizes and distances remains underexplored. In this work, we introduce a manually annotated benchmark, Q-Spatial Bench, with 271 questions across five categories designed for quantitative spatial reasoning and systematically investigate the performance of state-of-the-art vLMs on this task. Our analysis reveals that reasoning about distances between objects is particularly challenging for SoTA vLMs;however, some vLMs significantly outperform others, with an over 40-point gap between the two best performing models. We also make the surprising observation that the success rate of the top-performing vLM increases by 19 points when a reasoning path using a reference object emerges naturally in the response. Inspired by this observation, we develop a zero-shot prompting technique, SpatialPrompt, that encourages vLMs to answer quantitative spatial questions using reference objects as visual cues. By instructing vLMs to use reference objects in their reasoning paths via SpatialPrompt, Gemini 1.5 Pro, Gemini 1.5 Flash, and GPT-4v improve their success rates by over 40, 20, and 30 points, respectively. We emphasize that these significant improvements are obtained without needing more data, model architectural modifications, or fine-tuning. © 2024 Association for Computational Linguistics.

关键词： visual languages

来源：评论

学校读者我要写书评

暂无评论

Annotations in Different Steps of visual Analytics 16

Annotations in Different Steps of Visual Analytics

引用

16th International Joint conference on Computer vision, Imaging and Computer Graphics Theory and Applications (vISIGRAPP) / 12th International conference on Information visualization Theory and Applications (IvAPP)

作者： Schmidt, Christoph Grundel, Bastian Schumann, Heidrun Rosenthal, Paul Univ Rostock Inst Visual & Analyt Comp Rostock Germany Ernst Moritz Arndt Univ Greifswald Eye Care Ctr Greifswald Germany

ISBN: (纸本)9789897584886

Annotations in visual Analytics (vA) have become a common means to support the analysis by integrating additional information into the vA system. Here, annotations often differ between the individual steps of vA. For example, during data preprocessing it may be necessary to add information on the data, such as redundancy or discrepancy information, while annotations, used during exploration, often refer to the externalization of findings and insights. Describing the particular needs for these step-dependent annotations is challenging. To tackle this issue, we examine the data preprocessing, data cleansing, and data exploration steps for the analysis of heterogeneous and error prone data in respect to the design of specific annotations. By that, we describe their peculiarities for each step in the analysis, and thus aim to improve the visual analytics approach on clinical data. We show the applicability of our annotation concept by integrating it into an existing visual analytics tool to analyze and annotate data from the ophthalmic domain. In interviews and application sessions with experts, we assess the usefulness of our annotation concept for the analysis of the visual acuity development for patients, undergoing a specific therapy.

关键词： Annotation visual Analytics Heterogeneous Clinical data data Preprocessing data Cleansing data exploration

来源：评论

学校读者我要写书评

暂无评论

UOUO: Uncontextualized Uncommon Objects for Measuring Knowledge Horizons of vision Language Models

UOUO: Uncontextualized Uncommon Objects for Measuring Knowle...

引用

2024 conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Pi, Xinyu Wu, Mingyuan Jiang, Jize Zheng, Haozhen Tian, Beitong Zhai, Chengxiang Nahrstedt, Klara Hu, Zhiting University of California San Diego United States University of Illinois Urbana-Champaign United States

ISBN: (纸本)9798891761643

Smaller-scale vision-Language Models (vLMs) often claim to perform on par with larger models in general-domain visual grounding and question-answering benchmarks while offering advantages in computational efficiency and storage. However, their ability to handle rare objects, which fall into the long tail of data distributions, is less understood. To rigorously evaluate this aspect, we introduce the "Uncontextualized Uncommon Objects" (UOUO) benchmark. This benchmark focuses on systematically testing vLMs with both large and small parameter counts on rare and specialized objects. Our comprehensive analysis reveals that while smaller vLMs maintain competitive performance on common datasets, they significantly underperform on tasks involving uncommon objects. We also propose an advanced, scalable pipeline for data collection and cleaning, ensuring the UOUO benchmark provides high-quality, challenging instances. These findings highlight the need to consider long-tail distributions when assessing the true capabilities of vLMs. Code and project details for UOUO can be found at https://***/UOUOWebsite/. © 2024 Association for Computational Linguistics.

关键词： visual languages

来源：评论

学校读者我要写书评

暂无评论

Cloud Forensics analysis Framework for Secure and Efficient data Retrieval

Cloud Forensics Analysis Framework for Secure and Efficient ...

引用

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

作者： Revathi, v. Aishwarya, B.K. Tripathi, Manju Singh, Navdeep Alabdely, Haider Kumar, Ashwani New Horizon College of Engineering Department of Applied Sciences Bangalore India Institute of Aeronautical Engineering Hyderabad Dundigal India Ies College of Technology Department of Computer Science & Engineering M. P. Bhopal India Lovely Professional University Phagwara India The Islamic University Najaf Iraq Lloyd Institute of Engineering & Technology Knowledge Park II Uttar Pradesh Greater Noida201306 India

ISBN: (纸本)9798350373783

This study's Cloud Forensics analysis Framework seeks secure and effective data recovery methods. The framework's five strategies make digital evidence retrieval methodical and versatile. These include adaptive evidence collecting, secure data retrieval, machine learning-based anomaly detection, intelligent automation, and context-aware metadata analysis. Each approach targets a distinct forensic process step. This makes cloud forensics investigations seem difficult. The customizable Evidence gathering Algorithm adapts to cloud characteristics to collect evidence. This allows contextually aware and flexible digital evidence retrieval. The secure data Retrieval Techniques Algorithm protects data privacy and chain of custody with advanced encryption and validation. The Machine Learning-Based Anomaly Detection Algorithm checks protected data for anomalies before sending it, making the system safer. Investigators may focus on more complex cases by simplifying and sorting basic forensic activities with the Intelligent Automation Algorithm. The Context-Aware Metadata study Algorithm weights and prioritizes cloud metadata. This completes metadata analysis. Compared to other approaches, the framework has greater Precision, Recall, F1 Score, Processing Time, Resource Utilization, and Compatibility Score. Charts convey system operation in a simple manner. This powerful and adaptable Cloud Forensics analysis Framework improves cloud forensics by providing a thorough and effective solution to recover and safeguard data. The Cloud Forensics, This paper discusses context-aware metadata analysis, data retrieval, intelligent automation, machine learning-based anomaly detection, performance metrics, secure data retrieval techniques, support vector machines, and visual representation. © 2024 IEEE.

关键词： Cloud security

来源：评论

学校读者我要写书评

暂无评论

Applying visual analytics to firearms trafficking trails for the extraction of strategic intelligence

Applying visual analytics to firearms trafficking trails for...

引用

IEEE International conference on Big data

作者： E. Adamakis E. Sykianaki G. Margetis S. Ntoa C. Stephanidis Institute of Computer Science Foundation for Research and Technology Hellas Heraklion Greece Department of Computer Science University of Crete Heraklion Greece

ISBN: (数字)9798350362480

ISBN: (纸本)9798350362497

Big data analysis and insight extraction are critical in contemporary large-scale criminal investigations. Analyzing large sets of information related to criminal activities can assist in the identification of correlations between records which can be utilized as strategic intelligence information for the coordination of law enforcement efforts. A Law Enforcement Agency (LEA) officer tasked with this analysis requires tools that can enable Big data exploration while being intuitive and without requiring the need for prior knowledge. A common practice of data analysis software is the ability to create dashboards that can provide all the necessary visualizations and filters to support such an analysis. However, when it comes to dashboard-wide data exploration through drill-down or roll-up analysis; review and evaluation of correlations between data records; and creation of collections with information important to an investigation, current solutions fall short. In this paper, we present a system that addresses these limitations by facilitating the visualization and exploration of diverse data sources, allowing for individual or concurrent analysis via user-configured dashboards. Individual data records of interest to the analysis can be reviewed and their correlations evaluated by the user, allowing for the creation of collections of records and their relationships based on formalized types of correlations.

关键词： Correlation Filters data analysis Law enforcement Reviews visual analytics Soft sensors data visualization Big data data mining

来源：评论

学校读者我要写书评

暂无评论

The PEACE-Reviews dataset: Modeling Cognitive Appraisals in Emotion Text analysis

The PEACE-Reviews dataset: Modeling Cognitive Appraisals in ...

引用

conference on Empirical Methods in Natural Language Processing (EMNLP)

作者： Yeo, Gerard Christopher Jaidka, Kokil Natl Univ Singapore Inst Data Sci Singapore Singapore Natl Univ Singapore NUS Ctr Trusted Internet & Community Singapore Singapore

ISBN: (纸本)9798891760615

Cognitive appraisal plays a pivotal role in deciphering emotions. Recent studies have delved into its significance, yet the interplay between various forms of cognitive appraisal and specific emotions, such as joy and anger, remains an area of exploration in consumption contexts. Our research introduces the PEACE-Reviews dataset, a unique compilation of annotated autobiographical accounts where individuals detail their emotional and appraisal experiences during interactions with personally significant products or services. Focusing on the inherent variability in consumer experiences, this dataset offers an in-depth analysis of participants' psychological traits, their evaluative feedback on purchases, and the resultant emotions. Notably, the PEACE-Reviews dataset encompasses emotion, cognition, individual traits, and demographic data. We also introduce preliminary models that predict certain features based on the autobiographical narratives.(1).

关键词：

来源：评论

学校读者我要写书评

暂无评论

Enhancing Emotion Regulation Through virtual Reality Design Framework for Social-Emotional Learning (vRSEL)

引用

INTERNATIONAL JOURNAL OF ADvANCED COMPUTER SCIENCE AND APPLICATIONS 2024年第10期15卷 264-272页

作者： Hamzah, Irna Salwana, Ely Baghaei, Nilufar Billinghurst, Mark Arsad, Azhar Univ Kebangsaan Malaysia Inst Visual Informat Bangi Selangor Malaysia Univ Queensland Sch Elect Engn & Comp Sci St Lucia Australia Univ Auckland Auckland Bioengn Inst Auckland 1010 New Zealand V3X Malaysia Sdn Bhd 87AJalan Pulai 7Taman Pulai Skudai 81300 Johor Malaysia

virtual reality (vR) has swiftly progressed, transitioning from a niche technology primarily associated with gaming to a versatile tool with broad applications across entertainment, healthcare, education, and beyond. Social- emotional learning (SEL) is increasingly recognized for its role in enhancing individuals' social skills and emotional regulation. However, despite the growing body of research on vR, the development of specific design frameworks for integrating vR with SEL remains underexplored. This study addresses this gap by employing thematic analysis to identify the critical components necessary for a virtual Reality Design Framework for Social- Emotional Learning (vRSEL) aimed at improving emotion regulation among Malaysian adolescents. Through qualitative data derived from expert interviews in SEL and vR, this research proposes a framework that leverages immersive vR technology to create realistic, interactive scenarios that facilitate the practice and development of social-emotional skills. The framework emphasizes key design principles, including user interface (UI), presentation layer (PL), and brain activity (BA). Our findings suggest that vRSEL is a powerful tool for SEL, offering significant potential for educational environments. However, challenges such as technical barriers, content development, and educator training must be addressed to fully realize its benefits. This research highlights the promising role of vR in advancing SEL and lays the groundwork for further exploration and refinement of vRSEL in diverse educational settings.

关键词： virtual reality design framework social-emotional learning emotion regulation

来源：评论

学校读者我要写书评

暂无评论

A Comparative analysis of Image Partitioning and Compression Mechanisms 15

A Comparative Analysis of Image Partitioning and Compression...

引用

15th International conference on Advances in Computing, Control, and Telecommunication Technologies, ACT 2024

作者： Chaudhari, Sampada P. Thakur, Nileshsingh v. Dept. of Computer Science and Engineering Prof Ram Meghe College of Engineering & Management Badnera India Dept. of Computer Science and Engineering Yeshwantrao Chavan College of Engineering Nagpur India

ISBN: (纸本)9798331300579

This paper presents a comprehensive comparative analysis of image partitioning and compression mechanisms, two fundamental techniques in image processing and data compression. Image partitioning involves dividing an image into smaller components, while compression aims to reduce the size of digital images while preserving visual quality. The analysis encompasses various aspects of both techniques, including methodologies, algorithms, and performance metrics. For image partitioning, segmentation methods such as block-based, region-based, and hierarchical partitioning are examined, along with their respective advantages and limitations. Additionally, the paper explores compression mechanisms such as transform coding, quantization, entropy coding, and predictive coding, highlighting their roles in achieving efficient compression. Furthermore, the comparative analysis evaluates the tradeoffs between different partitioning and compression techniques in terms of compression ratio, image quality, computational complexity, and applicability across different types of images and applications. various compression standards such as JPEG and PNG are also reviewed to provide insights into real-world implementations. Finally, the paper discusses future research directions and emerging trends in image partitioning and compression, including the integration of machine learning techniques for improved segmentation and compression performance. The findings presented in this comparative analysis can serve as a valuable resource for researchers, practitioners, and developers working in the fields of image processing, data compression, and multimedia systems. © Grenze Scientific Society, 2024.

关键词： Image compression

来源：评论

学校读者我要写书评

暂无评论

AI Powered Ayurvedic Leaf Prediction and Disease Classification: A Comprehensive analysis

AI Powered Ayurvedic Leaf Prediction and Disease Classificat...

引用

2025 International conference on visual Analytics and data visualization, ICvADv 2025

作者： Sathya, R. Suganthi, M. Trisha, M. Janani, E. Dhivya, S. viswanathan, R.v. Kongunadu College of Engineering and Technology Department of It Tamil Nadu Trichy India

ISBN: (纸本)9798331521394

Ayurvedic medicine, a fundamental component of India's traditional healthcare framework, is heavily dependent on the quality and efficacy of medicinal plants. Nevertheless, the medicinal properties of these plants can be adversely affected by diseases stemming from pathogens and environmental factors. This research explores the utilization of artificial intelligence (AI) methodologies, such as deep learning (DL), machine learning (ML), and technologies, for the prompt identification and diagnosis of plant diseases. The review emphasizes recent developments in this field, evaluates their success in disease detection, and pinpoints areas where further research is needed. By merging AI technologies with traditional practices, the integrity of medicinal plants can be preserved, promoting sustainable agricultural practices and improved disease management © 2025 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Skin Lesion Detection Using Partial Differential Equations

Skin Lesion Detection Using Partial Differential Equations

引用

2025 International conference on visual Analytics and data visualization, ICvADv 2025

作者： Sree Lakshmi, D. Divya, A. Mugada, Gayathri Yadav Koduru, Tanusha Bingi, Mownika Madupu, Ishwarya Prasad v Potluri Siddhartha Institute of Technology Department of Computer Science and Engineering Vijayawada India

ISBN: (纸本)9798331521394

The technique used for image interest segmentation is known as segmentation. Processing pixels with matching attributes helps to accomplish this segmentation. Regions of this image become easier to process through part-based methods while avoiding complete image processing. The standard use of segmentation in pictures involves detecting objects as well as their edges. Technical image segmentation functions to break images into basic elements before creating a simplified understanding for easy analysis. A version of the image with reduced dimensions composed of uniform areas displaying similar visual patterns would serve as the best preference. Every image region must have a smooth intensity value that changes abruptly only at the region boundaries. The main segmentation approaches include one method for detecting original image objects and their contours. Edge detection systems produce matching results through the comparison of deformable curves with object contours by defining energy functional principles. © 2025 IEEE.

关键词： Wave equations

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：