检索结果-内蒙古大学图书馆

Increased Leverage of Transprecision Computing for machine vision applications at the Edge

JOURNAL OF SIGNAL processing SYSTEMS FOR SIGNAL image AND vIDEO TECHNOLOGY 2022年第10期94卷 1101-1118页

作者： Minhas, Umar Ibrahim Lee, JunKyu Mukhanov, Lev Karakonstantis, Georgios vandierendonck, Hans Woods, Roger Queens Univ Belfast Belfast Antrim North Ireland Queens Univ Belfast Inst Elect Commun & Informat Technol Belfast Antrim North Ireland Queens Univ Belfast Sch Elect Elect Engn & Comp Sci Belfast Antrim North Ireland Queens Univ Belfast Sch Elect Elect Engn & Comp Sci High Performance & Data Intens Comp Belfast Antrim North Ireland Queens Univ Belfast Inst Elect Commun & Informat Technol Ctr Data Sci & Scalable Comp Belfast Antrim North Ireland

The practical deployment of machine vision presents particular challenges for resource constrained edge devices. With a clear need to execute multiple tasks with variable workloads, there is a need for a robust approach that can dynamically adapt at runtime and which can maintain the maximum quality of service (QoS) within the available resource constraints. A lightweight approach that monitors the runtime workload constraints and leverages accuracy-throughput trade-offs on a graphics processing unit (GPU), is presented. It includes optimisation techniques that identify the configurations for each task in terms of optimal accuracy, energy and memory and management of the transparent switching between configurations. Using a neural network architecture search that statically generates a range of implementations that target a resource-precision trade-off, we explore the detection of the optimal parameters for the required QoS under specific memory and energy constraints. For an accuracy loss of 1%, we demonstrate that a 1.6x higher frame processing rate can be achieved on GPU with further improvements possible at further relaxed accuracy. In order to further improve the switching between configurations, we enhance the proposed mechanism by employing central processing units (CPUs) for offloading some of the executed frames, which helps to improve the frame rate by further 0.9%.

关键词： Edge Computing Approximate Computing Transprecision Computing machine vision

来源：评论

学校读者我要写书评

暂无评论

Automated stenosis detection in coronary artery disease using yolov9c: Enhanced efficiency and accuracy in real-time applications

引用

JOURNAL OF REAL-TIME image processing 2024年第5期21卷 177页

作者： Akgul, Muhammet Kozan, Hasan Ibrahim Akyurek, Hasan Ali Tasdemir, Sakir Necmettin Erbakan Univ Fac Med Konya Turkiye Necmettin Erbakan Univ Meram Vocat Sch Konya Turkiye Necmettin Erbakan Univ Fac Aeronaut & Astronaut Konya Turkiye Selcuk Univ Fac Technol Konya Turkiye

Coronary artery disease (CAD) is a prevalent cardiovascular condition and a leading cause of mortality. An accurate and timely diagnosis of CAD is crucial for treatment. This study aims to detect stenosis in real-time and automatically during angiographic imaging for CAD diagnosis, using the YOLOv9c model. A dataset comprising 8325 grayscale images was utilized, sourced from 100 patients diagnosed with one-vessel CAD. To enhance sensitivity and accuracy during the training, testing, and validation phases of stenosis detection, fine-tuning and augmentations were applied. The Python API, utilizing YOLO and Ultralytics libraries, was employed for these processes. The analysis revealed that the YOLOv9c model achieved remarkably high performance in both processing speed and detection accuracy, with an F1-score of 0.99 and mAP@50 of 0.99. The inference time was reduced to 18 ms, fine-tuning time to 3.5 h, and training time to 11 h. When the same dataset was tested using another significant diagnostic algorithm, SSD MobileNet v1, the YOLOv9c model outperformed it by achieving 1.36 x better F1-score and 1.42 x better mAP@50. These results indicate that the developed YOLOv9c algorithm can provide highly accurate and real-time results for stenosis detection.

关键词： machine learning YOLOv9c object detection Medical imaging Stenosis detection Coronary artery disease

来源：评论

学校读者我要写书评

暂无评论

Review on Deep Learning Network Architectures for image Reconstruction 5

Review on Deep Learning Network Architectures for Image Reco...

引用

5th International Conference on image processing and Capsule Networks, ICIPCN 2024

作者： Lokula, Babitha Prasad, L.v Narasimha Tirumuri, Ramakrishna K L deemed to be university Institute of Aeronautical Engineering department of ECE Vaddeshwaram India Institute of Aeronautical Engineering department of ECE Hyderabad India K L deemed to be university department of ECE Vaddeshwaram India

ISBN: (纸本)9798350367171

In order to obtain the low resolution(LR) image's detailed information, super resolution(SR) image reconstruction is essential. From 1D projections, we can reconstruct images in 2D and 3D. The LR images attained have different details of the same scene for which the super resolution reconstruction is possible. In the field of computer vision, image reconstruction is used in a variety of applications including robotics, entertainment, reverse engineering, augmented reality, human-computer interaction, and animation. In addition, applications in the medical field like Magnetic Resonance Imaging (MRI), Computed Tomography (CT), X-rays, and ultrasound, among others, make use of image reconstruction. We may encounter limitations while reconstructing the image, such as noise, missing information, a low signal to noise ratio, and a low contrast to noise ratio. We even come across different issues and challenges in application of SR techniques like image registration, computational efficiency, robustness and performance limitations. To address the above-mentioned issues, we study numerous SISR architectures. In this paper, we'll talk about various techniques, their benefits, drawbacks, and Single image Super Resolution (SISR) architectures that can help improve super resolution image reconstruction in a variety of applications. © 2024 IEEE.

关键词： image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Real-Time Farm Surveillance Using IoT and YOLOv8 for Animal Intrusion Detection

引用

FUTURE INTERNET 2025年第2期17卷 70-70页

作者： Delwar, Tahesin Samira Mukhopadhyay, Sayak Kumar, Akshay Singh, Mangal Lee, Yang-won Ryu, Jee-Youl Hosen, A. S. M. Sanwar Pukyong Natl Univ Dept Smart Robot Convergence & Applicat Engn Busan 48513 South Korea Symbiosis Int Deemed Univ Symbiosis Inst Technol Pune Campus Pune 412115 India Pukyong Natl Univ Dept Spatial Informat Engn Busan 48513 South Korea Pukyong Natl Univ Dept Informat & Commun Engn Busan 48513 South Korea Woosong Univ Dept Artificial Intelligence & Big Data Daejeon 34606 South Korea

This research proposes a ground-breaking technique for protecting agricultural fields against animal invasion, addressing a key challenge in the agriculture industry. The suggested system guarantees real-time intrusion detection and quick reactions by combining cutting-edge sensor technologies, image processing capabilities, and the Internet of Things (IoT), successfully safeguarding crops and reducing agricultural losses. This study involves a thorough examination of five models-Inception, Xception, vGG16, AlexNet, and Yolov8-against three different datasets. The Yolov8 model emerged as the most promising, with exceptional accuracy and precision, exceeding 99% in both categories. Following that, the Yolov8 model's performance was compared to previous study findings, confirming its excellent capabilities in terms of intrusion detection in agricultural settings. Using the capabilities of the Yolov8 model, an IoT device was designed to provide real-time intrusion alarms on farms. The ESP32cam module was used to build this gadget, which smoothly integrated this cutting-edge model to enable efficient farm security measures. The incorporation of this technology has the potential to transform farm monitoring by providing farmers with timely, actionable knowledge to prevent possible threats and protect agricultural production.

关键词： agricultural security animal detection computer vision crop protection early warning system IoT intrusion detection image processing machine learning sensor technology smart farming wildlife intrusion Yolo v7

来源：评论

学校读者我要写书评

暂无评论

Optimzied resnet model of convolutional neural network for under sea water object detection and classification

引用

MULTIMEDIA TOOLS AND applications 2023年第24期82卷 37551页

作者： Malathi, v. Manikandan, A. Krishnan, Kalimuthu Periyar Univ Dept Comp Sci Salem India Muthayammal Mem Coll Arts & Sci Dept Comp Sci Rasipuram India SRM Inst Sci & Technol Dept Elect & Commun Engn Chennai India

The world's ocean depths conceal a big mystery, and obtaining the information contained therein is a significant challenge that must be overcome. With the advent of computer vision technologies and robotics, the underwater environment is explored recently. The vast data collected from numerous underwater sensors have a variety of complications related to inadequate image quality, difficulty in acquiring training samples, and uncontrolled objects in the underwater environment. When these images are processed using machine learning techniques that involve manual intervention, the time taken to process a huge amount of images will be relatively high and prone to errors. To tackle these, we propose a novel hybrid capuchin-based coevolving particle swarm optimization ((HCPSO)-P-2) algorithm with a ResNet model of Convolutional Neural Network (CNN) architecture for underwater object identification. This work mainly aims to explore different underwater objects such as fish, corals, sea urchins, etc. The speckle-reducing anisotropic diffusion (SRAD) filter performs the pre-processing step. The denoising autoencoder (DA) is used for feature extraction which can enhance the partially distorted sample images and offer increased robustness. To overcome the overfitting issue in CNN, the (HCPSO)-P-2 algorithm is used. The experimental works are handled in MATLAB software. Both with and without pre-processing results in terms of SRAD filter are checked and evaluated. The proposed method's effectiveness is evaluated through various measures like accuracy, specificity, sensitivity, false-positive rate, false-negative rates, etc. The accuracy of the (HCPSO)-P-2-CNN classifier is higher when compared to the standard CNN classifier in recognizing the underwater objects when evaluated with different performance metrics.

关键词： ResNetCNN Underwater images Detection of objects SRAD filter Autoencoder Denoising Hybrid capuchin-based coevolving particle swarm optimization

来源：评论

学校读者我要写书评

暂无评论

Application of deep learning approaches for classification of diabetic retinopathy stages from fundus retinal images: a survey

引用

MULTIMEDIA TOOLS AND applications 2023年第14期83卷 43115页

作者： Mukherjee, N. Sengupta, S. Infosys Ltd Mysore India Aliah Univ Kolkata India

Diabetic retinopathy (DR) is an impediment of diabetes mellitus, which if not treated early may result in complete loss of vision, even without any preemptive symptoms. DR is caused by high level of glucose in the blood, causing alterations in the microvasculature of retina. However, early screening of diabetic patients through retinal fundus imaging, along with proper diagnosis and treatment can control the prevalence of DR complications. Manual inspection of pathological changes in retinal fundus images is an extremely challenging and tedious task. Therefore, computer-aided diagnosis (CAD) system is an efficient and effective method for early detection of DR and can greatly assist the ophthalmologists. CAD system encompasses DR detection and severity grading that includes detection, classification, localization and segmentation of lesions from the fundus images. Significant contributions have been made in DR severity grading using conventional image processing approaches using hand-engineered features and traditional machine-learning (ML) techniques. In the recent years, significant development of deep learning (DL) methods alleviated by the advancement of hardware computation power and efficient learning algorithms, has triumphed over the traditional ML methods in DR detection and grading tasks. Many researchers have employed the established as well as customized DL models in different DR image repositories and reported their findings. In this paper, we conduct a detailed review of the recent state-of-the-art contributions in the field of DL based DR classification by explaining their methodologies and highlighting their advantages and limitations. A detailed comparative study based on certain statistical parameters has also been conducted to quantitatively evaluate the methods, models and preprocessing techniques. In addition, the challenges in designing an efficient, accurate and robust deep-learning model for DR classification are explored in details to help t

关键词： Diabetic Retinopathy DR Stage Classification DR-related Lesions Medical image Analysis Computer-assisted Diagnosis machine Learning Deep Learning Survey

来源：评论

学校读者我要写书评

暂无评论

Research on taper thread’s compensation algorithm based on machine vision considering the inclined state effect and tooth profile distortion

引用

Multimedia Tools and applications 2023年第29期82卷 45983-46010页

作者： Lu, Qianhai Kong, Lingfei Tian, Dongzhuang Sun, Jin Li, Longlong Gong, Chunyuan School of Mechanical and Precision Instrument Engineering Xi’an University of Technology Shannxi Xi’an710048 China Xi’an Research Institute of China Coal Technology and Engineering Group Corp Shaanxi Xi’an710077 China School of Mechanical Engineering Xi’an Jiaotong University Shaanxi Xi’an710049 China

Drill pipe joint’s thread quality directly affects the machining performance and the drill pipe’s service life. machine vision can quickly detect thread parameters to determine the thread processing quality, but this method has low thread measurement accuracy due to factors such as drill pipe joint inclination and tooth shape distortion. This paper proposes a thread detection compensation algorithm based on thread geometry space and thread section projection theory to promote machine vision inspection accuracy. The distortion mechanism of thread section image caused by the drill pipe joint in an inclined state is revealed and a taper thread mathematical model is proposed. The difference equation is obtained by subtracting the projected contour from the theoretical contour, and the extreme value is obtained to correct the thread contour in the inclined state. Experiments show that the thread profile angle compensation efficiency can be increased by 60% under inclined conditions, and the requirements for the placement of drill pipe joints are reduced. A good agreement of the standard measurement with the experimental data proves the effectiveness of the proposed method. © 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.

关键词： Drill pipe

来源：评论

学校读者我要写书评

暂无评论

An Ensemble Technique for Classification of Oral Cancer by Using Histopathological Imaging 3rd

An Ensemble Technique for Classification of Oral Cancer by...

引用

3rd International Conference on machine vision and Augmented Intelligence, MAI 2023

作者： Saikia, Trishna Dhamaniya, Ashutosh Gupta, Puneet Singh, Koushlendra Kumar Department of Computer Science and Engineering IIT Indore Indore India Machine Vision and Intelligence Lab NIT Jamshedpur Jamshedpur India

ISBN: (纸本)9789819743582

Oral cancer primarily affects the oral chamber within the head and neck area, and underscores the critical need for effective classification to initiate timely treatment. Deep learning (DL)-based computer-aided diagnostic (CAD) systems have demonstrated notable success in various applications, offering accurate and prompt diagnosis of oral squamous cell carcinomas (OSCC). One of the challenges in biomedical image classification is the acquisition of a sufficiently large training dataset. Transfer learning presents an efficient solution by extracting general features from natural image datasets and adapting them to new image datasets. In this study, we focus on classifying OSCC histopathology images to develop a productive DL-based CAD solution. To this end, we employ an average weighted ensemble technique, harnessing the strengths of deep learning-based models. To address the limitation of a small dataset, we fine-tune pre-trained deep CNN models, specifically vGG-16, vGG-19, MobileNet-v2, and Inception-v3, within our proposed method. Additionally, we conduct a comprehensive comparative analysis of these models, considering classification accuracy, precision, recall, and F-score as metrics. Our experimental findings reveal that vGG-19 consistently delivers substantially superior performance compared to the other fine-tuned deep CNN models. However, our proposed weighted ensemble technique outperforms all these deep CNN models, particularly when employing the RMSProp optimizer. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Ensemble Histopathological imaging Inception-v3 MobileNet-v2 OSCC Transfer learning vGG-16 vGG-19

来源：评论

学校读者我要写书评

暂无评论

Artificial Intelligence Algorithms for Robotic Harvesting of Agricultural Produce 1

Artificial Intelligence Algorithms for Robotic Harvesting of...

引用

1st International Conference on AIML-applications for Engineering and Technology, ICAET 2025

作者： Kolhalkar, Nilesh R. Pandit, Anupama A. Kedar, Shridhar Ashok Yedukondalu, G. MKSSS's Cummins College of Engg. For Women Department of Mechanical Engineering MH Pune India Department of Computer Science & Engg. Pune MH Pune India Koneru Lakshmaiah Education Foundation Vaddeswaram Department of Mechanical Engineering A.P. Vijayawada India

ISBN: (纸本)9798350355611

Robotic harvesting of fruits and vegetables is an advanced technology that leverages Robotics, Artificial Intelligence, and machine vision to harvest the fruits autonomously from plants or trees. This technology aims to address labor shortages, enhance efficiency, reduce costs, and minimize damage to the fruit during harvesting. AI algorithms for fruit detection and harvesting are increasingly used in agricultural automation to improve efficiency and accuracy. The accuracy of detection algorithms in fruit detection and harvesting can differ reliant on various factors, including the type of algorithm used, the quality and diversity of the training data, the complexity of the environment, and the specific fruits being targeted. Advanced control algorithms integrated with image processing ensure that the robotic arm moves smoothly and accurately, minimizing the risk of bruising or damaging the fruit. Soft robotics and adaptive gripping technologies are discussed in the paper which can handle delicate fruits like grapes, without applying excessive force. machine vision integrated robot arm with novel gripper and cutter for harvesting cluster fruit like grapes is reported in the paper. Case studies of agricultural robots for Orchards, Greenhouses and Field Crops are discussed with detailed analysis along with challenges, future trends and innovations. © 2025 IEEE.

关键词： machine vision

来源：评论

学校读者我要写书评

暂无评论

Deep computer vision system for cocoa classification

引用

MULTIMEDIA TOOLS AND applications 2022年第28期81卷 41059-41077页

作者： Lopes, Jessica Fernandes Turrisi da Costa, victor G. Barbin, Douglas F. Pier Cruz-Tirado, Luis Jam Baeten, vincent Barbon Junior, Sylvio Londrina State Univ UEL Dept Elect Engn Londrina Parana Brazil Londrina State Univ UEL Dept Comp Sci Londrina Parana Brazil Univ Campinas UNICAMP Dept Food Engn Campinas Brazil Walloon Agr Res Ctr CRA W Gembloux Belgium Univ Trieste Dipartimento Ingn & Architettura DIA Trieste Italy

Cocoa hybridisation generates new varieties which are resistant to several plant diseases, but has individual chemical characteristics that affect chocolate production. image analysis is a useful method for visual discrimination of cocoa beans, while deep learning (DL) has emerged as the de facto technique for image processing . However, these algorithms require a large amount of data and careful tuning of hyperparameters. Since it is necessary to acquire a large number of images to encompass the wide range of agricultural products, in this paper, we compare a Deep Computer vision System (DCvS) and a traditional Computer vision System (CvS) to classify cocoa beans into different varieties. For DCvS, we used a Resnet18 and Resnet50 as backbone, while for CvS, we experimented traditional machine learning algorithms, Support vector machine (SvM), and Random Forest (RF). All the algorithms were selected since they provide good classification performance and their potential application for food classification A dataset with 1,239 samples was used to evaluate both systems. The best accuracy was 96.82% for DCvS (ResNet 18), compared to 85.71% obtained by the CvS using SvM. The essential handcrafted features were reported and discussed regarding their influence on cocoa bean classification. Class Activation Maps was applied to DCvS's predictions, providing a meaningful visualisation of the most important regions of the images in the model.

关键词： machine learning Deep learning Computer vision Food quality

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：