检索结果-内蒙古大学图书馆

Active Speaker detection Using Audio, Visual, and Depth Modalities: A Survey

IEEE ACCESS 2024年 12卷 96617-96634页

作者： Robi, Siti Nur Aisyah Mohd Ariffin, Muhammad Atiff Zakwan Mohd Izhar, Mohd Azri Mohd Ahmad, Norulhusna Kaidi, Hazilah Mad UTM Fac Artificial Intelligence Kuala Lumpur 54100 Malaysia

The rapid progress of multimodal signal processing in recent years has cleared the way for novel applications in human-computer interaction, surveillance, and telecommunication. Active Speaker detection (ASD) is a critical pre-processing step with numerous applications such as voice recognition, speaker diarization, and noise reduction. This paper comprehensively reviews ASD, including various ASD methods and datasets based on these three modalities - audio, visual and/or depth modalities. ASD methods are broadly categorised into two categories: single modality ASD and multi-modality ASD. This review looks at the most common ASD modalities, which include audio-based ASD (A-ASD), visual-based ASD (V-ASD), audio-visual ASD (AV-ASD), and audio-visual-depth ASD (AVD-ASD). Each strategy is well-detailed, including model-based and neural network-based approaches. Finally, the challenges and future research opportunities are highlighted in order to expand its broader use.

关键词： Visualization Speech recognition Reviews Cameras Accuracy Surveys Microphones detection algorithms Audio-visual systems Active speaker detection voice activity detection multi-modalities audio-visual processing deep learning

来源：评论

学校读者我要写书评

暂无评论

C2I-CAT: Class-to-Image Cross Attention Transformer for Out-of-Distribution detection

引用

IEEE ACCESS 2024年 12卷 62793-62803页

作者： Chung, Jaeho Cho, Seokho Choi, Hyunjun Jo, Daeung Jung, Yoonho Choi, Jin Young Seoul Natl Univ Automation & Syst Res Inst ASRI Dept Elect & Comp Engn Seoul 08826 South Korea Medintech Seoul 03100 South Korea Samsung Adv Inst Technol SAIT Samsung Elect Suwon 16678 South Korea

In our work, we have empirically found that Vision Transformer (ViT) could not extract object-centric features when applied to out-of-distribution (OOD) detection. To make object-centric attention, we design an additional module that employs a cross-attention between class-wise token proxy and feature token sequence of an input image. For inference suitable to our cross-attention structure with multiple class-wise token proxies, we propose a score ensemble that can be applied to any scoring function. Compared to ViT, the proposed inference scheme achieves outperforming performance by synergizing with our cross-attention structure. Through experiments, we demonstrate that the proposed cross-attention structure with score ensemble inference improves largely near OOD detection performance, where FPR95 improvement in near OOD detection compared to the state-of-the-art method becomes 2.55% for CIFAR-10 and 2.67% for CIFAR-100, keeping competitive classification accuracy.

关键词： Feature extraction Transformers Training Remote sensing Sampling methods Data mining detection algorithms Near out-of-distribution (OOD) detection vision transformer class-wise cross attention

来源：评论

学校读者我要写书评

暂无评论

Lightweight detection Model RM-LFPN-YOLO for Rebar Counting

引用

IEEE ACCESS 2024年 12卷 3936-3947页

作者： Liu, Haodong Cheng, Wansheng Li, Chunwei Xu, Yaowen Fan, Song Univ Sci & Technol Liaoning Sch Elect & Informat Engn Anshan 114051 Liaoning Peoples R China

In this study, we propose a novel lightweight detection model for rebar counting, which is rectified mobilenet lightweight feature pyramid network based on YOLO (RM-LFPN-YOLO). The model incorporates a lightweight backbone network that integrates the coordinate attention (CA) mechanism, a lightweight feature pyramid network (LFPN), and a loss function that combines focal loss and efficient intersection over union (EIOU) loss, all meticulously designed to enhance the model's performance. Experimental results demonstrate that our improved algorithm, with a mere 25.08M parameters, computes efficiently at 7.60G with an input size of 416 pixels. Additionally, it achieves an impressive average precision (AP) of 99.03% at an IOU of 0.5. The proposed lightweight model can be deployed on embedded devices and achieve efficient rebar detection and counting performance.

关键词： Feature extraction Convolutional neural networks Computational modeling Steel Bars YOLO Task analysis detection algorithms attention LFPN focal loss

来源：评论

学校读者我要写书评

暂无评论

Enhancing DDoS Attack detection and Mitigation in SDN Using an Ensemble Online Machine Learning Model

引用

IEEE ACCESS 2024年 12卷 51630-51649页

作者： Alashhab, Abdussalam Ahmed Zahid, Mohd Soperi Isyaku, Babangida Elnour, Asma Abbas Nagmeldin, Wamda Abdelmaboud, Abdelzahir Abdullah, Talal Ali Ahmed Maiwada, Umar Danjuma Alasmarya Islamic Univ Fac Informat Technol Dept Comp Sci Zliten Libya Univ Teknol PETRONAS Dept Comp & Informat Sci Seri Iskandar 32610 Malaysia Sule Lamido Univ Fac Comp & Informat Technol Dept Comp Sci Kano 700271 Jigawa Nigeria King Khalid Univ Appl Coll Comp Dept Girls Sect Muhayel 62529 Aseer Saudi Arabia Prince Sattam Bin Abdulaziz Univ Coll Comp Engn & Sci Dept Informat Syst Al Kharj 11942 Saudi Arabia Sultan Qaboos Univ Humanities Res Ctr Seeb Oman

Software Defined Networks (SDN) offer dynamic reconfigurability and scalability, revolutionizing traditional networking. However, countering Distributed Denial of Service (DDoS) attacks remains a formidable challenge for both traditional and SDN-based networks. The integration of Machine Learning (ML) into SDN holds promise for addressing these threats. While recent research demonstrates ML's accuracy in distinguishing legitimate from malicious traffic, it faces difficulties in handling emerging, low-rate, and zero-day DDoS attacks due to limited feature scope for training. The ever-evolving DDoS landscape, driven by new protocols, necessitates continuous ML model retraining. In response to these challenges, we propose an ensemble online machine-learning model designed to enhance DDoS detection and mitigation. This approach utilizes online learning to adapt the model with expected attack patterns. The model is trained and evaluated using SDN simulation (Mininet and Ryu). Its dynamic feature selection capability overcomes conventional limitations, resulting in improved accuracy across diverse DDoS attack types. Experimental results demonstrate a remarkable 99.2% detection rate, outperforming comparable models on our custom dataset as well as various benchmark datasets, including CICDDoS2019, InSDN, and slow-read-DDoS. Moreover, the proposed model undergoes comparison with industry-standard commercial solutions. This work establishes a strong foundation for proactive DDoS threat identification and mitigation in SDN environments, reinforcing network security against evolving cyber risks.

关键词： Adaptation models Denial-of-service attack Machine learning Computer crime Data models Software defined networking Ensemble learning detection algorithms Scalability DDoS attacks LDDoS attacks SDN OML ensemble detection and mitigation

来源：评论

学校读者我要写书评

暂无评论

Lightweight YOLOv8 for Wheat Head detection

引用

IEEE ACCESS 2024年 12卷 66214-66222页

作者： Fang, Chen Yang, Xiang Guilin Univ Technol Coll Comp Sci & Engn Guilin 541006 Guangxi Peoples R China

Accurate real-time observations of wheat head growth are crucial for effective agricultural management. However, the dense distribution of wheat heads often leads to severe overlap in imagery. Existing target detection algorithms face challenges in overcoming this problem, rendering them ineffective for real-time field computations using portable devices. Therefore, this study proposes a lightweight you-only-look-once (YOLO) model with a simplified structure and a more powerful attention mechanism. A limitation of the traditional YOLO model is its complex structure, as it requires a substantial number of parameters and its accuracy is unsatisfactory. We remove the modules designed for large targets and reduce the number of detection heads from three to two. Moreover, we add an improved feature pyramid network to the neck, resulting in improved parameter count and accuracy over traditional YOLO methods. To improve inferencing, we replace the spatial pyramid pooling (SPP) module with a simplified SPP-fast type. Finally, a large separable kernel attention and wise intersection-over-union method are introduced to integrate the attention mechanisms, and we replace the loss function to improve the discriminative capabilities of the model. Experimental results on the Global Wheat Head Dataset demonstrate a 53% reduction in memory usage, 27% decrease in computational load, and 5.2 frames-per-second increase in detection speed over extant methods. The proposed model also achieves 3.9%, 2.1%, and 1.3% improvements in terms of precision, recall, and mean average precision, respectively, and it exhibits lightweight and portable characteristics. Overall, the proposed method has fast inference speed and effectively alleviates the problem of low recognition of dense wheat head.

关键词： Feature extraction YOLO Computational modeling Magnetic heads Image color analysis Data models Object detection Crops Agriculture detection algorithms Attention mechanism lightweight model object detection wheat head

来源：评论

学校读者我要写书评

暂无评论

Automated detection of Spinal Lesions From CT Scans via Deep Transfer Learning

引用

IEEE ACCESS 2024年 12卷 65310-65322页

作者： Camisa, Andrea Montanari, Giovanni Testa, Andrea Falzetti, Luigi Avnet, Sofia Baldini, Nicola Notarstefano, Giuseppe Helsing London W1T 3BL England Salcef Grp I-00138 Rome Italy Univ Bologna Dept Elect Elect & Informat Engn I-40126 Bologna Italy IRCCS Ist Ortoped Rizzoli Biomed Sci & Technol & Nanobiotechnol Lab I-40136 Bologna Italy Univ Bologna Dept Biomed & Neuromotor Sci I-40126 Bologna Italy

Convolutional Neural Networks are being increasingly applied to the detection of anomalies in Computed Tomographies (CTs). The goal of this paper is to implement an automated Computer-Aided detection (CADe) system for spinal lesions using CTs and Convolutional Neural Networks pre-trained on commercial datasets. The proposed pipeline works as follows. The CADe takes in input CT scans and is equipped with an intuitive GUI to allow physicians to use it as a support tool for diagnoses. From the CT scans, the CADe selects volumes containing the vertebrae and extracts 2D slices of these volumes. These slices are pre-processed and then analyzed using a VGG19 Convolutional Neural Network and a tailored, binary classifier. The neural network identifies healthy vertebrae and vertebrae containing lesions (e.g. metastases, primary tumors, lytic and sclerotic lesions). For training and testing purposes, we generated a dataset from CTs of vertebrae from patients treated at IRCCS Rizzoli Orthopaedics Institute of Bologna, Italy, between 2009 and 2019. Both healthy and lesioned vertebrae retrieved with different tomography machines and setups are considered. The dataset has been enlarged using data augmentation techniques and subsequently used to train a wide range of deep learning models. We perform an in-depth benchmark study to assess the performance of the considered classifier against different Convolutional Neural Networks pre-trained on the ImageNet dataset and exploiting Transfer Learning techniques. Leveraging Transfer Learning techniques, we reached 93.43% accuracy and a recall of 92.99%.

关键词： Lesions Computed tomography Graphical user interfaces Transfer learning Magnetic resonance imaging Deep learning X-rays detection algorithms Neural networks CT scan deep learning lesion detection neural networks spinal lesions transfer learning

来源：评论

学校读者我要写书评

暂无评论

Overcoming the Challenges of Long-Tail Distribution in Nighttime Vehicle detection

引用

IEEE INTELLIGENT SYSTEMS 2024年第2期39卷 51-60页

作者： Zhang, Houwang Chan, Leanne Lai Hang City Univ Hong Kong Dept Elect Engn Hong Kong 999077 Peoples R China City Univ Hong Kong Elect Engn Hong Kong 999077 Peoples R China

As the basic task of an intelligent transportation system, nighttime vehicle detection is associated with many challenges. Existing methods usually ignore the significant challenges that arise from imbalanced class distribution among vehicles, which always leads to poor detection for vehicles that belong to tail classes. By analyzing the existing solutions for long-tail object detection and considering the complex and diverse characteristics of nighttime traffic scenarios, we propose an enhanced detection approach based on anomaly detection. In addition, to tackle disturbance from complex lights, we reconstruct the loss function for background proposals, thus allowing the detector to pay more attention to hard-classified proposals and learn to distinguish vehicle lights from disturbed light resources. Comprehensive experiments prove that, compared with generic approaches, our proposed method can effectively solve the problem of long-tail distribution in nighttime vehicle detection and improve robustness in complex environments.

关键词： Proposals Vehicle detection Training Automobiles Urban areas Intelligent transportation systems Vehicle safety Road safety Road traffic control detection algorithms

来源：评论

学校读者我要写书评

暂无评论

Guided Attention and Joint Loss for Infrared Dim Small Target detection

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2024年 62卷

作者： Tong, Yunfei Liu, Jing Fu, Zhiling Wang, Zhe Yang, Hai Niu, Saisai Tan, Qinyan East China Univ Sci & Technol Minist Educ Key Lab Smart Mfg Energy Chem Proc Shanghai 200237 Peoples R China East China Univ Sci & Technol Dept Comp Sci & Engn Shanghai 200237 Peoples R China China Aerosp Sci & Technol Corp Shanghai Aerosp Control Technol Inst Shanghai 201109 Peoples R China China Aerosp Sci & Technol Corp Res & Dev Ctr Infrared Detect Technol Shanghai 201109 Peoples R China

Infrared dim small target (IDST) detection is of great significance in security surveillance and disaster relief. However, the complex background interference and tiny targets in infrared images keep it still a long-term challenge. Existing deep learning models stack network layers to expand the model fitting capability, but this operation also increases redundant features which reduce model speed and accuracy. Meanwhile, small targets are more susceptible to positional bias, with this dramatically reducing the model's localization accuracy. In this article, we propose a guided attention and joint loss (GA-JL) network for infrared small target detection. More specifically, the method visualizes the feature maps at each resolution through a two-branch detection head (TDH) module, filters out the features that are strongly related to the task, and cuts out the redundant features. On this basis, the guided attention (GA) module guides the prediction layer features using the features that are associated closely with the task and combines spatial and channel bidirectional attention to make the prediction layer feature maps embedded with effective messages. Finally, through the joint loss (JL) module, the target position regression is performed with multiangle metrics for enhancing the target detection accuracy. Experimental results of our method on the SIATD, SIRST, and IRSTD_1k datasets reveal that it is capable of accurately identifying IDSTs, remarkably reduces the false alarm rate, and outperforms other methods.

关键词： Feature extraction Object detection Accuracy Attention mechanisms Sensitivity Geoscience and remote sensing Fitting detection algorithms Deep learning Windows Guided-attention (GA) mechanism infrared dim small target (IDST) detection joint loss (JL) positional bias YOLO series

来源：评论

学校读者我要写书评

暂无评论

BotFence: A Framework for Network-Enriched Botnet detection and Response With SmartNICs

引用

IEEE ACCESS 2024年 12卷 114878-114893页

作者： Seo, Hyunmin Shin, Seungwon Lee, Seungsoo Korea Adv Inst Sci & Technol Sch Elect Engn Daejeon 34141 South Korea Incheon Natl Univ Dept Comp Sci & Engn Incheon 22012 South Korea

The scale of botnet attacks is on the rise, yet traditional network security systems are inadequate to effectively respond to these threats, primarily due to high false positive rates and the extensive manpower required for analysis. In contrast, the cutting-edge method of intrusion detection, known as provenance-based analysis, offers a novel paradigm by establishing causality between host events for meticulous examination. Nonetheless, this method faces challenges in analyzing the payload of network packets, which contains critical attack information resides, due to performance efficiency constraints from packet inspection. To address these challenges, we introduce BotFence, a pioneering approach that integrates payload inspection of network packets with provenance-based analysis to enhance botnet intrusion detection and response. Notably, our system leverages SmartNICs to minimize the impact on network performance. Our system initially gathers and analyzes events within the host system, representing them as Tactics, Techniques, and Procedures (TTP). Concurrently, it collects and scrutinizes the network packets associated with these events, integrating the relationships between these TTPs and the collected network data into a Network-enhanced Threat Provenance Graph (NTPG) model that we devised. Consequently, our system provides a comprehensive security analysis of the network with minimal overhead. Demonstrations with complex attack scenarios show that BotFence successfully identifies and mitigates automated botnet infection in real time, analyzing more than 99. 9% host events in 1 ms, without degrading network performance.

关键词： Botnet Payloads Servers Inspection Real-time systems Object recognition detection algorithms Network security endpoint detection and response programmable dataplane network security

来源：评论

学校读者我要写书评

暂无评论

Using Graph Attention Networks in Healthcare Provider Fraud detection

引用

IEEE ACCESS 2024年 12卷 132786-132800页

作者： Mardani, Shahla Moradi, Hadi Univ Tehran Coll Engn Sch Elect & Comp Engn Tehran 111554563 Iran SKKU Intelligent Syst Res Inst Suwon 16419 South Korea

Healthcare fraud increases healthcare expenses for insurers, premiums for policyholders, and dissatisfaction of legitimate patients and causes severe damage to the health system. Therefore, it is critically important to address healthcare fraud detection. Most fraud detection models consider only claims data for analysis. Since committing healthcare fraud can include more than one party, i.e., healthcare providers, physicians, and patients, it is crucial to consider the relationship among them. In this paper, we propose a healthcare provider fraud detection model that applies the effect of the parties' interdependencies on the claims data. It leverages a graph attention network for embedding the relationships and classifies samples using a feed-forward neural network. Our explicit contribution is using the latent information of the interdependency of claims' parties to detect healthcare provider fraud more accurately. The information, along with other features, can identify complex patterns of fraud. We tested our approach on the healthcare provider fraud detection dataset and reached 0.56 recall compared to best available approaches such as GTN with 0.5 and XGBoost with 0.46 recall.

关键词： Medical services Fraud Feature extraction Vectors Support vector machines Logistics detection algorithms Insurance Graphical models Fraud detection graph attention network graph embedding healthcare

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：