检索结果-内蒙古大学图书馆

SwimVG: Step-wise Multimodal Fusion and Adaption for Visual Grounding

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Shi, Liangtao Liu, Ting Hu, Xiantao Hu, Yue Yin, Quanjun Hong, Richang Key Laboratory of Knowledge Engineering with Big Data Hefei University of Technology Hefei230009 China Ministry of Education and School of Computer Science and Information Engineering Hefei University of Technology Hefei230009 China School of systems engineering National University of Defense Technology Hunan Province Changsha410073 China Department of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210014 China

Visual grounding aims to ground an image region through natural language, which heavily relies on cross-modal alignment. Most existing methods transfer visual/linguistic knowledge separately by fully fine-tuning uni-modal pre-trained models, followed by a simple stack of visual-language transformers for multimodal fusion. However, these approaches not only limit adequate interaction between visual and linguistic contexts, but also incur significant computational costs. Therefore, to address these issues, we explore a step-wise multimodal fusion and adaption framework, namely SwimVG. Specifically, SwimVG proposes step-wise multimodal prompts (Swip) and cross-modal interactive adapters (CIA) for visual grounding, replacing the cumbersome transformer stacks for multimodal fusion. Swip can improve the alignment between the vision and language representations step by step, in a token-level fusion manner. In addition, weight-level CIA further promotes multimodal fusion by cross-modal interaction. Swip and CIA are both parameter-efficient paradigms, and they fuse the cross-modal features from shallow to deep layers gradually. Experimental results on four widely-used benchmarks demonstrate that SwimVG achieves remarkable abilities and considerable benefits in terms of efficiency. Our code is available at https://***/liuting20/SwimVG. Copyright © 2025, The Authors. All rights reserved.

关键词： Visual languages

Towards Harmless Rawlsian Fairness Regardless of Demographic Prior

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Xuanqian Li, Jing Tsang, Ivor W. Ong, Yew-Soon School of Computer Science and Engineering Beihang University China Institute of High-Performance Computing Agency for Science Technology and Research Singapore Centre for Frontier AI Research Agency for Science Technology and Research Singapore College of Computing and Data Science Nanyang Technological University Singapore

Due to privacy and security concerns, recent advancements in group fairness advocate for model training regardless of demographic information. However, most methods still require prior knowledge of demographics. In this study, we explore the potential for achieving fairness without compromising its utility when no prior demographics are provided to the training set, namely harmless Rawlsian fairness. We ascertain that such a fairness requirement with no prior demographic information essential promotes training losses to exhibit a Dirac delta distribution. To this end, we propose a simple but effective method named VFair to minimize the variance of training losses inside the optimal set of empirical losses. This problem is then optimized by a tailored dynamic update approach that operates in both loss and gradient dimensions, directing the model towards relatively fairer solutions while preserving its intact utility. Our experimental findings indicate that regression tasks, which are relatively unexplored from literature, can achieve significant fairness improvement through VFair regardless of any prior, whereas classification tasks usually do not because of their quantized utility measurements. The implementation of our method is publicly available at https://***/wxqpxw/VFair. © 2024, CC BY.

关键词： Differential privacy

A Dual-Level Cancelable Framework for Palmprint Verification and Hack-Proof data Storage

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yang, Ziyuan Kang, Ming Teoh, Andrew Beng Jin Gao, Chengrui Chen, Wen Zhang, Bob Zhang, Yi The College of Computer Science Key Laboratory of Data Protection and Intelligent Management Ministry of Education Sichuan University Chengdu610065 China The School of Cyber Science and Engineering Sichuan University Chengdu610065 China The School of Electrical and Electronic Engineering College of Engineering Yonsei University Seoul Korea Republic of The College of Computer Science Sichuan University Chengdu610065 China The Pattern Analysis and Machine Intelligence Group Department of Computer and Information Science University of Macau Taipa China The School of Cyber Science and Engineering Key Laboratory of Data Protection and Intelligent Management Ministry of Education Sichuan University Chengdu610065 China

In recent years, palmprints have been widely used for individual verification. The rich privacy information in palmprint data necessitates its protection to ensure security and privacy without sacrificing system performance. Existing systems often use cancelable technologies to protect templates, but these technologies ignore the potential risk of data leakage. Upon breaching the system and gaining access to the stored database, a hacker could easily manipulate the stored templates, compromising the security of the verification system. To address this issue, we propose a dual-level cancelable palmprint verification framework in this paper. Specifically, the raw template is initially encrypted using a competition hashing network with a first-level token, facilitating the end-to-end generation of cancelable templates. Different from previous works, the protected template undergoes further encryption to differentiate the second-level protected template from the first-level one. The system specifically creates a negative database (NDB) with the second-level token for dual-level protection during the enrollment stage. Reversing the NDB is NP-hard and a fine-grained algorithm for NDB generation is introduced to manage the noise and specified bits. During the verification stage, we propose an NDB matching algorithm based on matrix operation to accelerate the matching process of previous NDB methods caused by dictionary-based matching rules. This approach circumvents the need to store templates identical to those utilized for verification, reducing the risk of potential data leakage. Extensive experiments conducted on public palmprint datasets have confirmed the effectiveness and generality of the proposed framework. Upon acceptance of the paper, the code will be accessible at https://***/Deep-Imaging-Group/NPR. Copyright © 2024, The Authors. All rights reserved.

关键词： Digital storage

Toward Cost-effective Adaptive Random Testing: An Approximate Nearest Neighbor Approach

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huang, Rubing Cui, Chenhui Lian, Junlong Towey, Dave Sun, Weifeng Chen, Haibo The School of Computer Science and Engineering Macau University of Science and Technology Taipa999078 China The School of Computer Science and Communication Engineering Jiangsu University Jiangsu Zhenjiang212013 China The School of Computer Science University of Nottingham Ningbo China Zhejiang Ningbo315100 China The School of Big Data and Software Engineering Chongqing University Chongqing401331 China

Adaptive Random Testing (ART) enhances the testing effectiveness (including fault-detection capability) of Random Testing (RT) by increasing the diversity of the random test cases throughout the input domain. Many ART algorithms have been investigated such as Fixed-Size-Candidate-Set ART (FSCS) and Restricted Random Testing (RRT), and have been widely used in many practical applications. Despite its popularity, ART suffers from the problem of high computational costs during test-case generation, especially as the number of test cases increases. Although several strategies have been proposed to enhance the ART testing efficiency, such as the forgetting strategy and the k-dimensional tree strategy, these algorithms still face some challenges, including: (1) Although these algorithms can reduce the computation time, their execution costs are still very high, especially when the number of test cases is large;and (2) To achieve low computational costs, they may sacrifice some fault-detection capability. In this paper, we propose an approach based on Approximate Nearest Neighbors (ANNs), called Locality-Sensitive Hashing ART (LSH-ART). When calculating distances among different test inputs, LSH-ART identifies the approximate (not necessarily exact) nearest neighbors for candidates in an efficient way. LSH-ART attempts to balance ART testing effectiveness and efficiency. © 2023, CC BY.

关键词： Nearest neighbor search

Efficient DNA Sequence Identification: A Classification Approach Integrating Gaussian RBM 1

学校读者我要写书评

暂无评论

Efficient DNA Sequence Identification: A Classification Appr...

1st IEEE International Conference on Electronics, Computing, Communication and Control Technology, ICECCC 2024

作者： Gulati, Chahat Barman, Biswajit Balde, Govind Hanmantrao Arul Mary Rexy, V. Sachdeva, Meena Patil, Harshal Birla Institute of Technology and Science BITS Department of Electronics and Communication Rajasthan Pilani India Asian International University Department of Biotechnology Imphal West Manipur India N.T.V.S'S G.T.Patil Arts Commerce and Science College Department of Zoology Nandurbar India Saveetha College of Liberal Arts and Sciences SIMATS Department of Commerce Chennai India JIMS Engineering Management and Technical Campus Department of Computer Science Greater Noida India Sri Balaji University Department of Data Science Maharashtra Pune India

ISBN: (纸本)9798350371802

Using machine learning techniques, this article examines cancer patients' DNA sequences. Cancer is certainly one of the most common and devastating diseases in the contemporary era, with a large number of new cases and deaths every year. Many people lose their lives each year to this terrible illness. Methods to combat this illness are the focus of intense investigation by scientists from a wide range of fields. Recently, machine learning has improved the analysis of cancer patients' DNA sequences. Sequencing is crucial for preprocessing, feature extraction, and model training. Among the preprocessing stages is cleaning up the sequence, which entails eliminating PCR repetitions, impurities, and vectors. Feature extraction has been carried out on each DNA sequence in order to retrieve statistical characteristics. It is of utmost importance to regulate features accurately when training a Gaussian RBM. A grade of 96.44% demonstrated a considerable improvement in accuracy, according to the results. © 2024 IEEE.

关键词： Polymerase chain reaction

Robust UAV Policy Learning for Urban Infrastructure Surface Screening*

学校读者我要写书评

暂无评论

Robust UAV Policy Learning for Urban Infrastructure Surface ...

International Conference on Advanced Robotics and Mechatronics (ICARM)

作者： Bingqing Du Uddin Md. Borhan Tianxing Chen Jianyong Chen Jianqiang Li Jie Chen College of Computer Science and Software Engineering Shenzhen University Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China

ISBN: (数字)9798350385724

ISBN: (纸本)9798350385731

Unmanned Aerial Vehicles (UAVs) have emerged as pivotal tools for infrastructure inspection, mitigating risks to human life and enhancing operational efficiency. Despite advancements, achieving autonomous surface coverage inspection in GPS-denied environments remains a significant challenge. This paper presents a novel approach to enhance UAV policy learning for surface screening of urban infrastructure in environments lacking GPS. Utilizing Deep Reinforcement Learning (DRL), the methodology enables UAVs to autonomously conduct surface coverage inspections more robustly and efficiently. A key innovation is the Optimal Speed Decision (OSD) model, which allows UAVs to adjust flight speed dynamically based on scene perception, enhancing efficiency and safety. Additionally, an Off-Policy Critic-Based Reward Shaping (OPCBRS) method is introduced to improve learning efficiency and performance by addressing the issue of sparse rewards using offline data. Simulation results confirm the effectiveness of these methods, with the OSD model demonstrating significant improvements in navigating complex scenarios and maintaining a balance between efficiency and robustness over traditional uniform speed policies. This research offers a robust framework for UAV-based infrastructure inspection in GPS-denied environments, advancing autonomous UAV applications.

关键词： Technological innovation Navigation Simulation Inspection Autonomous aerial vehicles Deep reinforcement learning Robustness Real-time systems Safety Robots

INFOCON: CONCEPT DISCOVERY WITH GENERATIVE AND DISCRIMINATIVE INFORMATIVENESS

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Liu, Ruizhe Luo, Qian Yang, Yanchao HKU Musketeers Foundation Institute of Data Science The University of Hong Kong Hong Kong School of Electronics Engineering and Computer Science Peking University China Department of Electrical and Electronic Engineering The University of Hong Kong Hong Kong

We focus on the self-supervised discovery of manipulation concepts that can be adapted and reassembled to address various robotic tasks. We propose that the decision to conceptualize a physical procedure should not depend on how we name it (semantics) but rather on the significance of the informativeness in its representation regarding the low-level physical state and state changes. We model manipulation concepts - discrete symbols - as generative and discriminative goals and derive metrics that can autonomously link them to meaningful sub-trajectories from noisy, unlabeled demonstrations. Specifically, we employ a trainable codebook containing encodings (concepts) capable of synthesizing the end-state of a sub-trajectory given the current state - generative informativeness. Moreover, the encoding corresponding to a particular sub-trajectory should differentiate the state within and outside it and confidently predict the subsequent action based on the gradient of its discriminative score - discriminative informativeness. These metrics, which do not rely on human annotation, can be seamlessly integrated into a VQ-VAE framework, enabling the partitioning of demonstrations into semantically consistent sub-trajectories, fulfilling the purpose of discovering manipulation concepts and the corresponding sub-goal (key) states. We evaluate the effectiveness of the learned concepts by training policies that utilize them as guidance, demonstrating superior performance compared to other baselines. Additionally, our discovered manipulation concepts compare favorably to human-annotated ones while saving much manual effort. Our code is available at: https://***/InfoCon/. © 2024, CC BY.

关键词： Trajectories

Enhanced Synthetic Images Classification with Vision Transformers and Effective data Augmentation Techniques

学校读者我要写书评

暂无评论

Enhanced Synthetic Images Classification with Vision Transfo...

IEEE International Conference on Multi Topic

作者： Khadija Haider Muhammad Sohail Abbas Asif Muhammad Muhid Qaiser Department of Artificial Intelligence and Data Science National University of Computer and Emerging Sciences Islamabad Pakistan Department of Software Engineering National University of Computer and Emerging Sciences Islamabad Pakistan

ISBN: (数字)9798331507213

ISBN: (纸本)9798331507220

Synthetic images have made significant advancements, resulting in highly realistic and high-quality visuals generated at an unprecedented speed. However, the proliferation of synthetic images raises growing concerns regarding their potential misuse, ethical implications, and associated risks in various sectors. This study focuses on leveraging data augmentation techniques to enhance the accuracy and reliability of synthetic image detection. In our study, we fine-tuned a Vision Transformer (ViT) on the CIFAKE dataset, comprising real and synthetic images, augmented using various techniques such as transformations, photometric adjustments, geometric changes, and noise addition. The best results were obtained with ViT fine-tuned by distortions and noise augmentation, achieving an accuracy rate of 98.66%, demonstrating that sophisticated augmentation techniques improve performance. data augmentation plays a crucial role in handling the increasing realism and sophistication of synthetic images, contributing to more effective content moderation, digital forensics efforts, and the mitigation of potential risks associated with synthetic media.

关键词： Performance evaluation computer vision Visualization Accuracy Computational modeling Transformers Distortion data augmentation Reliability Image classification

Diabetes Prediction Using Machine Learning Techniques: A Case Study from Jordan

学校读者我要写书评

暂无评论

Diabetes Prediction Using Machine Learning Techniques: A Cas...

Information Technology (ACIT)

作者： Raed Alazaidah Radwan Batyha Ghassan Samara Esraa Al Henawi Maram Y. Al-Safarini Ahmad A. Samhan Ali Elrashidi Mo’Ath Alluwaici Deana Al-Omari Department of Data Science and AI Faculty IT Zarqa University Zarqa Jordan Department of Computer Science Faculty of IT Applied Science Private University Amman Jordan Department of Computer Science Faculty of IT Zarqa University Zarqa Jordan Department of Software Engineering Faculty of IT Zarqa University Zarqa Jordan Department of Electrical Engineering University of Business and Technology Jeddah Saudi Arabia

ISBN: (数字)9798331540012

ISBN: (纸本)9798331540029

Diabetes is one of the most common diseases in Jordan. It is the main reason of death among Jordanian adult citizens. Worldwide, 48% of all deaths are due to Diabetes occurred before the age of 70 years. Hence, this research is interested in the early prediction of this disease among Jordanian people. To achieve this main objective, Machine Learning (ML) is utilized through a large number of classification models and considering five well-known evaluation metrics. These classification models have been trained on a primary dataset that has been collected for the purpose of the research. The results revealed that BayesNet and NaiveBayes are the best classification models to handle the task of the early prediction of Diabetes in Jordan

关键词： Measurement Machine learning Predictive models Diabetes Bayes methods Information technology Diseases