检索结果-内蒙古大学图书馆

Date Fruit detection and Classification Based on Its Variety Using Deep Learning Technology

IEEE ACCESS 2024年 12卷 190666-190677页

作者： Almutairi, Ahad Alharbi, Jawza Alharbi, Shouq Alhasson, Haifa F. Alharbi, Shuaa S. Habib, Shabana Qassim Univ Coll Comp Dept Informat Technol Buraydah 51452 Saudi Arabia

There are many types of date fruits, each with similar physical properties and slight differences in color, shape, and fleshiness. Due to the many attributes to consider, distinguishing between the different types of dates can be challenging and time consuming. Classifying and sorting the post-harvest fully ripe (Rutab) dates red into edible dates, is an essential stage of the date fruit production industries. Date production industries in Saudi Arabia still perform this process manually, which may result in reduced productivity and an increase in the date fruit waste. The purpose of this study was to use Deep Learning technology to detect and classify them by using the You Only Look Once (YOLO) algorithm. Different YOLO models (YOLOv5, YOLOv7, YOLOv8) have been trained using 1735 images with 9 types of date fruits. The accuracy of this study has been investigated using the metrics of Recall, Precision, and F1-score metrics. Based on the experimental results, YOLOv8 achieves a mean Recall of 0.99 % and Precision of 0.991% with an Intersection over Union (IoU) colorred in the range of [0-1], a mean Average Precision of 0.994%, a mean Average Precision of [0.5:0.95] and a mean Average Precision of [0.5:0.85]. Results show that the used model can accurately detect and classify date fruits based on their surface quality, enhancing the productivity as a result of multiple harvests due to uneven ripening of the fruit.

关键词： YOLO Training Accuracy Measurement Deep learning Data models Sorting Artificial neural networks detection algorithms Machine learning Performance evaluation Artificial neural network deep learning detection performance image analysis machine learning models

来源：评论

学校读者我要写书评

暂无评论

Multi Pattern Features-Based Spoofing detection Mechanism Using One Class Learning

引用

IEEE ACCESS 2024年 12卷 117523-117540页

作者： Ustubioglu, Beste Tahaoglu, Gul Ustubioglu, Arda Ulutas, Guzin Amerini, Irene Kilic, Muhammed Karadeniz Tech Univ Comp Engn Dept TR-61080 Trabzon Turkiye Univ Florence Dept Informat Engn I-50139 Florence Italy Trabzon Univ Dept Management Informat Syst TR-61335 Trabzon Turkiye Sapienza Univ Rome Dept Comp Control & Management Engn I-00185 Rome Italy

Automatic Speaker Verification systems are prone to various voice spoofing attacks such as replays, voice conversion (VC) and speech synthesis. Malicious users can perform specific tasks such as controlling the bank account of someone, taking control of a smart home, and similar activities, by using advanced audio manipulation techniques. This study presents a Multi-Pattern Features Based Spoofing detection mechanism using the modified ResNet architecture and OC-Softmax layer to detect various LA and PA spoofing attacks. We proposed a novel Pattern features-based audio spoof detection scheme. The scheme contains three branches to evaluate different patterns on a Mel spectrogram of the audio file. This is the first work for the audio spoofing detection task using three different pattern representations of Mel spectrogram with modified ResNet architecture and OC-Softmax layer. Through the proposed network, we can extract pattern images from the Mel spectrogram and gives each of them into modified ResNet architecture. At the last step of each network, we use OC-Softmax to obtain a score for the current pattern image and then the method fuses three scores to label the input audio. Experimental results on the ASVspoof 2019 and ASVspoof 2021 corpuses show that the proposed method achieves better results in the challenges of ASVspoof 2019 than state-of-the-art methods. For example, in the logical access scenario, our model improves the tandem decision cost function and equal error rate scores by 0.06% and 2.14%, respectively, compared with state-of-the-art methods. Additionally, experiments illustrate that the proposed fused decision improved the performance of the system.

关键词： Feature extraction Transforms Spectrogram Training Cost function Convolutional neural networks Deepfakes Forgery detection algorithms Deep fake audio audio forgery forgery detection

来源：评论

学校读者我要写书评

暂无评论

Quickest Change detection for Unnormalized Statistical Models

引用

IEEE TRANSACTIONS ON INFORMATION THEORY 2024年第2期70卷 1220-1232页

作者： Wu, Suya Diao, Enmao Banerjee, Taposh Ding, Jie Tarokh, Vahid Duke Univ Dept Elect & Comp Engn Durham NC 27708 USA Univ Pittsburgh Dept Ind Engn Pittsburgh PA 15213 USA Univ Minnesota Sch Stat Minneapolis MN 55455 USA

Classical quickest change detection algorithms require modeling pre-change and post-change distributions. Such an approach may not be feasible for various machine learning models because of the complexity of computing the explicit distributions. Additionally, these methods may suffer from a lack of robustness to model mismatch and noise. This paper develops a new variant of the classical Cumulative Sum (CUSUM) algorithm for the quickest change detection. This variant is based on Fisher divergence and the Hyvarinen score and is called the Hyvarinen score-based CUSUM (SCUSUM) algorithm. The SCUSUM algorithm allows the applications of change detection for unnormalized statistical models, i.e., models for which the probability density function contains an unknown normalization constant. The asymptotic optimality of the proposed algorithm is investigated by deriving expressions for average detection delay and the mean running time to a false alarm. Numerical results are provided to demonstrate the performance of the proposed algorithm.

关键词： Change detection algorithms Delays Computational modeling detection algorithms Machine learning Random variables Numerical models Quickest change detection CUSUM fisher divergence score matching unnormalized models

来源：评论

学校读者我要写书评

暂无评论

Shared-Memory Parallel algorithms for Community detection in Dynamic Graphs

Shared-Memory Parallel Algorithms for Community Detection in...

引用

IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW)

作者： Subhajit Sahu Kishore Kothapalli Dip Sankar Banerjee International Institute of Information Technology Hyderabad India Hyderabad India Indian Institute of Technology Jodhpur India Karwar Rajasthan India

ISBN: (数字)9798350364606

ISBN: (纸本)9798350364613

Community detection is the problem of identifying natural divisions in networks. A relevant challenge in this problem is to find communities on rapidly evolving graphs. In this paper, we present our parallel Dynamic Frontier (DF) approach. Given a batch update of edge deletions or insertions, this approach incrementally identifies an approximate set of affected vertices in the graph with minimal overhead. We apply this approach to both Louvain, a high quality, and Label Propagation Algorithm (LPA), a fast static community detection algorithm. Our approach achieves a mean speedup of 7.3 × and 6.7 ×, when applied to Louvain and LPA respectively, compared to our parallel and optimized implementation of Δ-screening, a recently proposed state-of-the-art approach. Finally, we show how to combine Louvain and LPA with the DF approach to arrive at a hybrid algorithm. This algorithm produces high-quality communities while providing a speedup of 2.0 x on top of DF -based Louvain.

关键词： Distributed processing Heuristic algorithms Image edge detection Conferences Approximation algorithms detection algorithms Parallel algorithms

来源：评论

学校读者我要写书评

暂无评论

Efficient Tumor detection and Classification Model Based on ViT in an End-to-End Architecture

引用

IEEE ACCESS 2024年 12卷 106096-106106页

作者： Huang, Ning-Yuan Liu, Chang-Xu Nanjing Tech Univ Coll Artificial Intelligence Coll Comp & Informat Engn Nanjing 211816 Peoples R China

Accurate tumor detection and classification are crucial for cancer diagnosis and treatment. Traditional medical image analysis methods face many challenges when dealing with highly heterogeneous tumor images, such as large differences in image quality and unclear or complex tumor features. Although breakthroughs have been made in image processing with deep learning techniques, there are still limitations in identifying small or irregular tumors. Existing tumor detection models often rely on local feature extraction, neglecting global information and subtle differences in the images, which limits their accuracy and robustness in practical applications. To address these issues, this paper proposes a deep learning model that integrates Feature Pyramid Network (FPN) and Vision Transformer (ViT) within an end-to-end architecture. Firstly, the model extracts rich features at multiple scales through FPN, covering various aspects from cellular structures to tissue layouts. Then, by introducing ViT, the model can effectively process and analyze global features, particularly achieving higher accuracy in recognizing ambiguous or complex tumor patterns. The self-attention mechanism further enhances the model's focus on critical regions of the image, improving its ability to detect subtle differences. Finally, the design of the end-to-end architecture enhances the overall efficiency and consistency of the model, facilitating global optimization and further improving detection and classification performance. The experimental results show that compared to existing techniques, this model demonstrates higher recognition accuracy on medical image datasets such as TCIA, BraTS, LUNA, and Camelyon17. The accuracy and F1 scores improved by 4.65% to 6.24%. These algorithmic improvements not only enhance the efficiency and accuracy of tumor detection but also provide new pathways for the application of deep learning in medical image analysis.

关键词： Tumors Medical diagnostic imaging Accuracy Image segmentation Transformers Lesions Task analysis detection algorithms Biomedical imaging Feature pyramid network vision transformer self-attention mechanism tumor detection medical image

来源：评论

学校读者我要写书评

暂无评论

Applied AI in Defect detection for Additive Manufacturing: Current Literature, Metrics, Datasets, and Open Challenges

引用

IEEE INSTRUMENTATION & MEASUREMENT MAGAZINE 2024年第4期27卷 46-53页

作者： Djenouri, Youcef Srivastava, Gautam Lin, Jerry Chun-Wei Univ South Eastern Norway Kongsberg Norway Norwegian Res Ctr Oslo Norway IDEAS NCBR Warsaw Poland Brandon Univ Brandon MB Canada Silesian Tech Univ Gliwice Poland

Defect detection in additive manufacturing refers to the evaluation of collected industrial images and the identification of parts that cause anomalies to optimize decision-making in an industrial production context. The advent of the Internet of Things and the widespread installation of electronic sensors, such as image sensors in industrial production lines, have expanded the processing capabilities of analytics tools. By extracting visual information via convolutional operations, deep learning-based algorithms have recently overcome drawbacks of traditional machine learning methods. This paper provides an analysis of contemporary defect detection techniques based on deep learning. Existing methods for defect detection algorithms in additive manufacturing are discussed. In terms of potential research to improve defect detection in additive manufacturing, the difficulties and emerging trends related to defect detection through deep learning are also outlined.

关键词： Deep learning Measurement Image sensors Productivity Visualization Machine learning algorithms Three-dimensional printing detection algorithms Performance evaluation Industrial engineering Defect detection Artificial intelligence Data models

来源：评论

学校读者我要写书评

暂无评论

Improved YOLOv5 Based Deep Learning System for Jellyfish detection

引用

IEEE ACCESS 2024年 12卷 87838-87849页

作者： Pham, Thi-Ngot Nguyen, Viet-Hoan Kwon, Ki-Ryong Kim, Jae-Hwan Huh, Jun-Ho Natl Korea Maritime & Ocean Univ Dept Data Informat Busan 49112 South Korea Natl Korea Maritime & Ocean Univ Interdisciplinary Major Ocean Renewable Energy Eng Busan 49112 South Korea Pukyong Natl Univ Dept Artificial Intelligence Convergence Busan 48513 South Korea Intown Co Ltd Busan 46207 South Korea Natl Korea Maritime & Ocean Univ Dept Data Sci Busan 49112 South Korea

Massive jellyfish outbreaks have put human lives and marine ecosystems in great danger. As a result, the jellyfish detection methods have drawn a lot of attention, following two directions optical and sonar imaging. This work focuses on using optical imagery and CNN-based deep-learning object detection models to detect jellyfish. While labeled data of jellyfish play an important part in training deep learning models, there are a few open and available labeled datasets. Hence, we create our dataset to train these models using our model-assisted labeling method with over 11 thousand images of underwater jellyfish and corresponding annotation files in PASCAL VOC format. Our model-assisted labeling method saves the work of classical manual labeling by 70 percent, which is developed into application with YOLOv5. However, the YOLOv5 baseline suffers from the trade-off between real-time performance and low accuracy. Hence, an improved YOLOv5-nano is introduced based on adding GAM and replacing conventional Conv with CoordCov modules into the backbone of the conventional structure. The experiment results show that our improved model increases the accuracy of the conventional one by 1.3% and outperforms others including RetinaNet, SSD, Faster R-CNN, YOLOv6, and YOLOv8 at 89.1% mAP@0.5. On generalization performance, we verify the effectiveness of our work by conducting a test set of 15 different types of jellyfish with various shapes, colors, resolutions, and backgrounds. To conclude, our work establishes a comprehensive system from labeling the data, improving object detectors, and developing a feasible real-time jellyfish detector.

关键词： Jellyfish YOLO Annotations Optical imaging Detectors Biological system modeling Oceans Deep learning detection algorithms Aquaculture Jellyfish detection deep learning YOLOv5 coordinate attention GAM CoordCov

来源：评论

学校读者我要写书评

暂无评论

An Analysis of GPS Spoofing Attack and Efficient Approach to Spoofing detection in PX4

引用

IEEE ACCESS 2024年 12卷 46668-46677页

作者： Jung, Ji Hyuk Hong, Mi Yeon Choi, Hyeongjun Yoon, Ji Won Korea Univ Sch Cybersecur Seoul 02841 South Korea

Unmanned Aerial Vehicles (UAVs) are aerial vehicles that can go to a particular position without human control or with remote human control. It is because unmanned control mainly relies on position estimation that a lot of research has been studied on GPS spoofing attacks. detection methods against GPS spoofing attacks mainly include monitoring RF signals or IMU sensor values inside UAVs. In this work, we analyze GPS attack and detection in an advanced autopilot system, PX4 without excluding RF detection. Recent studies have shown that GPS spoofing is unable to evade the detection using EKF sensor fusion. Therefore, this paper experiment whether the detection could be evaded by strengthening the attacker model. This paper classifies attacker models according to whether an attacker knows the true position of UAVs or the estimated position by UAV and proposed attack methods depending on each model in PX4. We inject GPS value which our attack method intends to UAV during mission flight in simulation environment. By manipulating the GPS driver code of PX4 controller, we inject GPS value the attack method wants into UAV in physical environment. Finally, we observed the innovation test ratio during GPS spoofing and demonstrate that GPS spoofing attack is possible keeping the innovation test ratio under the anomaly detection criterion. After that, we proposed our approach which scales the EKF's Kalman gain randomly to increase the detection method's efficiency and experiment with the attack methods. This detection method has little effect on the test ratio without the attack. But during the attack, the innovation test ratio was increased higher than the anomaly detection criterion so that the attack could be detected.

关键词： detection algorithms Position measurement detection GPS spoofing attack position estimation PX4 UAV

来源：评论

学校读者我要写书评

暂无评论

A Robust Framework for Distributional Shift detection Under Sample-Bias

引用

IEEE ACCESS 2024年 12卷 59598-59611页

作者： Torpmann-Hagen, Birk Riegler, Michael A. Halvorsen, Pal Johansen, Dag UiT Arctic Univ Tromso Inst Comp Sci N-9019 Tromso Norway SimulaMet Dept Holist Syst N-0167 Oslo Norway

Deep Neural Networks have been shown to perform poorly or even fail altogether when deployed in real-world settings, despite exhibiting excellent performance on initial benchmarks. This typically occurs due to relative changes in the nature of the production data, often referred to as distributional shifts. In an attempt to increase the transparency, trustworthiness, and overall utility of deep learning systems, a growing body of work has been dedicated to developing distributional shift detectors. As part of our work, we investigate distributional shift detectors that utilize statistical tests of neural network-based representations of data. We show that these methods are prone to fail under sample-bias, which we argue is unavoidable in most practical machine learning systems. To mitigate this, we implement a novel distributional shift detection framework which explicitly accounts for sample-bias via a simple sample-selection procedure. In particular, we show that the effect of sample-bias can be significantly reduced by performing statistical tests against the most similar data in the training set, as opposed to the training set as a whole. We find that this improves the stability and accuracy of a variety of distributional shift detection methods on both covariate- and semantic-shifts, with improvements to balanced accuracy typically ranging between 0.1 and 0.2, and false-positive-rates often being eliminated altogether under bias.

关键词： Training Data models Training data Detectors Deep learning System performance Distributed databases detection algorithms system support neural networks distributional shift OOD detection sample bias statistical tests

来源：评论

学校读者我要写书评

暂无评论

Cyberbullying detection and Abuser Profile Identification on Social Media for Roman Urdu

引用

IEEE ACCESS 2024年 12卷 123339-123351页

作者： Atif, Ayesha Zafar, Amna Wasim, Muhammad Waheed, Talha Ali, Amjad Ali, Hazrat Shah, Zubair Univ Engn & Technol Lahore Dept Comp Sci Lahore 39161 Pakistan Univ Management & Technol Sialkot Campus Lahore 51310 Pakistan Hamad Bin Khalifa Univ HBKU Qatar Fdn Coll Sci & Engn CSE Div Informat & Comp Technol Doha Qatar Univ Stirling Comp Sci & Math Stirling FK9 4LA Scotland

In today's digital era, the escalating phenomenon of cyberbullying is a pervasive and growing concern. With the increasing prevalence of social media platforms, such as Twitter, online abusive behavior has become a significant issue that often leads to unpleasant experiences for users. Manual detection of abnormal and bullying behavior within the realm of social media is inherently not scalable. Moreover, most existing studies on cyberbullying detection have been predominantly conducted in English and very limited work has been done on Urdu (a widely used language in Asia). This paper presents an approach for detecting cyberbullying in Roman Urdu tweets and identifying abuser profiles on Twitter. Firstly, we develop a text corpus of Roman Urdu tweets with user profile data. Subsequently, we employ Gated Recurrent Unit (GRU) model coupled with the application of word2vec technique for word embedding to develop a cyberbullying detection model. Furthermore, we present temporal abusive tweet probability analysis method to provide a nuanced analysis of the number of bullying and non-bullying tweets sent by individuals within a specific time interval. To evaluate the performance, we compare the GRU-based approach with other machine learning models. The results show that the GRU model with lexical normalization gives the best results with an accuracy of 97% and F1-measure of 97%.

关键词： Cyberbullying Hate speech Blogs Support vector machines Feature extraction Deep learning Data models detection algorithms Social networking (online) Machine learning Identification of persons Cyberbullying detection social media Roman Urdu machine learning deep learning abuser profile identification

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：