检索结果-内蒙古大学图书馆

A computer-aided diagnostic system for mammograms based on YOLOv3

MULTIMEDIA TOOLS AND APPLICATIONS 2022年第14期81卷 19257-19281页

作者： Zhao, Jianhui Chen, Tianquan Cai, Bo Wuhan Univ Sch Comp Sci Wuhan Hubei Peoples R China Wuhan Univ Sch Cyber Sci & Engn Wuhan Hubei Peoples R China

Due to a large amount of noise in medical images, the task of detecting and classifying the lesions of mammograms remains a huge challenge. Based on the existing deep learning methods, focusing on the diversity of breast cancer lesion types, this paper proposes a computer-aided diagnosis system based on YOLOv3 (You Only Look Once version 3) convolutional neural network for mammograms. In this system, we integrate detection and multi-classification problems of breast lesions into a regression problem, thereby simultaneously accomplish the two tasks in one framework. The proposed computer-aided diagnosis system is mainly divided into three components: preprocessing part of the original mammograms, deep convolutional neural network based on YOLOv3, processing and evaluation of the network output. We use the dataset from CBIS-DDSM to train three models: general model, mass model and microcalcification model. These trained models can detect the position of the input mammograms in different situations, and then classify them into mass, microcalcification, benign, malignant, and other categories. After evaluating the performance by using test set images, the accuracy rates of the general model, mass model, and microcalcification model trained by our system reach 93.667 %, 97.767 %, 96.870 % in the detection task, and 93.927 %, 98.121 %, 97.045 % in the classification task. The computer-aided diagnosis system performs well in lesion detection and classification tasks with high-noise mammograms, reflecting well robustness.

关键词： Breast cancer Deep learning Convolutional neural network object detection and classification YOLOv3

来源：评论

学校读者我要写书评

暂无评论

An evolving ensemble model of multi-stream convolutional neural networks for human action recognition in still images

引用

NEURAL COMPUTING & APPLICATIONS 2022年第11期34卷 9205-9231页

作者： Slade, Sam Zhang, Li Yu, Yonghong Lim, Chee Peng Northumbria Univ Fac Engn & Environm Dept Comp & Informat Sci Newcastle Upon Tyne NE1 8ST Tyne & Wear England Royal Holloway Univ London Dept Comp Sci Surrey TW20 0EX England Nanjing Univ Posts & Telecommun Coll Tongda Nanjing Peoples R China Deakin Univ Inst Intelligent Syst Res & Innovat Waurn Ponds Vic 3216 Australia

Still image human action recognition (HAR) is a challenging problem owing to limited sources of information and large intra-class and small inter-class variations which requires highly discriminative features. Transfer learning offers the necessary capabilities in producing such features by preserving prior knowledge while learning new representations. However, optimally identifying dynamic numbers of re-trainable layers in the transfer learning process poses a challenge. In this study, we aim to automate the process of optimal configuration identification. Specifically, we propose a novel particle swarm optimisation (PSO) variant, denoted as EnvPSO, for optimal hyper-parameter selection in the transfer learning process with respect to HAR tasks with still images. It incorporates Gaussian fitness surface prediction and exponential search coefficients to overcome stagnation. It optimises the learning rate, batch size, and number of re-trained layers of a pre-trained convolutional neural network (CNN). To overcome bias of single optimised networks, an ensemble model with three optimised CNN streams is introduced. The first and second streams employ raw images and segmentation masks yielded by mask R-CNN as inputs, while the third stream fuses a pair of networks with raw image and saliency maps as inputs, respectively. The final prediction results are obtained by computing the average of class predictions from all three streams. By leveraging differences between learned representations within optimised streams, our ensemble model outperforms counterparts devised by PSO and other state-of-the-art methods for HAR. In addition, evaluated using diverse artificial landscape functions, EnvPSO performs better than other search methods with statistically significant difference in performance.

关键词： Convolutional neural network Ensemble model Human action recognition Hyper-parameter optimisation object detection and classification

来源：评论

学校读者我要写书评

暂无评论

Vision-Based Humanoid Robotic Palm for Amputee Person using Deep Learning 6

Vision-Based Humanoid Robotic Palm for Amputee Person using ...

引用

6th IEEE International Conference on Emerging Smart Computing and Informatics, ESCI 2024

作者： Shilaskar, Swati Bhatlawande, Shripad Bagmare, Sandesh Shejole, Prem Vishwakarma Institute of Technology Pune Department of Electronics and Telecommunication Engineering Pune411037 India

ISBN: (纸本)9798350306613

Amputees often have to face certain challenges due to their physical limitations. This naturally affects their work and have to encounter issues of unemployment. Overcoming people's perspective of an amputee not working is something that the proposed system wishes to change. Thus, making the person self reliable and independent. This paper introduces a Humanoid Robotic Palm designed to assist amputees, enhancing their independence and quality of life. YOLOv5, VGG16, and ResNet architectures of CNN are used to train the model using a custom dataset of 2,667 annotated images featuring common objects, and that model is integrated with robotic palm for real-time object detection and classification. The Humanoid Robotic Palm is programmed to grasp the object by forming a grip according to the detected object. The Robotic Palm is controlled using Arduino Uno. PyFirmata Library is used to connect the Deep learning model with the Arduino Uno microcontroller. The accuracy achieved is 98.92% using ResNet Architecture, 96.55% using VGG16 Architecture, and 84.90% using YOLOv5. © 2024 IEEE.

关键词： Arduino Uno Gripping object detection and classification Pyfirmata Roboflow

来源：评论

学校读者我要写书评

暂无评论

Web-Based Traffic-Sign detection 23

Web-Based Traffic-Sign Detection

引用

31st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM SIGSPATIAL GIS)

作者： Pedersen, Kasper F. Torp, Kristian Aalborg Univ Dept Comp Sci Aalborg Denmark

ISBN: (纸本)9798400701689

Detecting traffic signs on images has many applications within the transportation area, e.g., speed limit detection, navigation, asset management, and autonomous vehicles. A significant challenge in detecting traffic signs is that a large set of labeled images are needed to train, test, and validate an object detector. In this paper, we demonstrate a webapp that enables traffic-sign detection of 169 common traffic signs and downloads the results. Users can contribute to improving object detection by (1) annotating user-uploaded images, (2) verifying traffic signs detected on existing imagery, or (3) donating new imagery by uploading images, e.g., taken with smartphones. The set of detected traffic signs is stepwise increased by correcting mistakes in object detection using the online verification part of the system. The user can immediately download all verified traffic-sign objects. Users can also download 43,995 traffic-sign objects from 378 different classes. This existing set of traffic-sign objects is updated nightly with objects two or more users have verified. The demonstration includes detecting traffic signs on imagery uploaded by the audience and downloading the traffic sign detected, e.g., to be used internally in an organization.

关键词： annotation imagery object detection and classification

来源：评论

学校读者我要写书评

暂无评论

Utilizing synthetic images to enhance the automated recognition of small-sized construction tools

引用

AUTOMATION IN CONSTRUCTION 2024年 163卷

作者： Han, Soeun Park, Wonjun Jeong, Kyumin Hong, Taehoon Koo, Choongwan Incheon Natl Univ Div Architecture & Urban Design Incheon 22012 South Korea Yonsei Univ Dept Architecture & Architectural Engn Seoul 03722 South Korea

Previous studies on vision-based classifiers often overlooked the need for detecting small-sized construction tools. Considering the substantial variations in these tools' size and shape, it is essential to train models using synthetic images that encompass diverse angles and distances. This study aimed to improve the performance of classifiers for small-sized construction tools by leveraging synthetic data. Three classifiers were proposed using YOLOv8 algorithm, varying in data composition: (i) 'Real -4000 ' : 4000 authentic images;(ii) 'Hybrid -4000 ' : 2000 authentic and 2000 synthetic images;(iii) 'Hybrid -8000 ' : 4000 authentic and 4000 synthetic images. To assess practical applicability, a test dataset of 144 samples for each type was collected directly from construction sites. Results revealed that the 'Hybrid -8000 ' model, utilizing synthetic images, excelled at 94.8% of mAP_0.5. This represented a significant 15.2% improvement, affirming its practical applicability. These classifiers hold promise for enhancing safety and advancing real -time automation and robotics in construction.

关键词： Small-sized construction tools object detection and classification Synthetic images Practical applicability Construction site

来源：评论

学校读者我要写书评

暂无评论

COMPUTER VISION AND MACHINE LEARNING TO CREATE AN ADVANCED PICK-AND-PLACE ROBOTIC OPERATION USING INDUSTRY 4.0 TRENDS

COMPUTER VISION AND MACHINE LEARNING TO CREATE AN ADVANCED P...

引用

ASME International Mechanical Engineering Congress and Exposition (IMECE)

作者： Guerra-Zubiaga, David Franklin, Angelicia Escobar-Escobar, Diego Lemley, Timothey Hariri, Neeyaz Plattel, Jeremy Ham, Chan Kennesaw State Univ Dept Robot & Mechatron Engn Marietta GA 30144 USA

ISBN: (纸本)9780791886632

This paper explores integrating several Industry 4.0 trends within a Kawasaki Robot and Vanderlande intelligent manufacturing execution system located at Kennesaw State University (KSU) in the United States of America. Several of the key Industry 4.0 trends that will be discussed within this paper include, but are not limited to, the following topics: Machine Learning (ML), Supervisory Control and Data Acquisition ( SCADA), Industrial Internet of Things (IIoT), and Cloud Manufacturing (CM). Several researchers explored these Industry 4.0 trends in manufacturing operations, but very few of them researched intelligent robotics grippers using MES and implementing advanced computer vision technologies. This research scopes in this direction. The research novelty contribution relies on exploring advanced intelligent robotic grippers while providing some scenarios to understand the next generation of automation systems according to Industry 4.0 trends by implementing both computer vision (CV) and machine learning (ML) aspects through an MES.

关键词： Intelligent Robot Gripper Smart Manufacturing Automation Industry 4.0 Trends Robotic Machine Learning Computer Vision object detection and classification

来源：评论

学校读者我要写书评

暂无评论

Super Resolution for Augmented Reality Applications

Super Resolution for Augmented Reality Applications

引用

41st IEEE Conference on Computer Communications (IEEE INFOCOM)

作者： Li, Vladislav Amponis, George Nebel, Jean-Christophe Argyriou, Vasileios Lagkas, Thomas Ouzounidis, Savvas Sarigiannidis, Panagiotis Kingston Univ Fac Sci Engn & Comp London England K3Y Ltd Res & Dev Dept Sofia Bulgaria Int Hellen Univ Dept Comp Sci Kavala Campus Kavala Greece South East European Res Ctr Thessaloniki Greece Univ Western Macedonia Dept Elect & Comp Engn Kozani Greece

ISBN: (纸本)9781665409261

Latest developments in machine learning (ML), adversarial networks, combined with increasingly powerful IoT devices via the introduction of efficient processors, are bringing about the implementation of near real-time object detection and classification for augmented reality (AR) and virtual reality (VR) applications. This paper intends to explore new object detection and classification technologies leveraging super-resolution (SR), that have the potential to be integrated into small, mobile and low-power AR/VR devices. SR in conjunction with novel object detection and classification algorithms are examined in the presented paper, with the ultimate goal of proposing a low-footprint Generative Adversarial Network (GAN)-based framework capable of receiving an LR input and outputting an SR-supported recognition model based on FRRCNN, YOLOv3 or Retina.

关键词： object detection and classification Super-Resolution AR/VR applications

来源：评论

学校读者我要写书评

暂无评论

A New Benchmark Model for the Automated detection and classification of a Wide Range of Heavy Construction Equipment

引用

JOURNAL OF MANAGEMENT IN ENGINEERING 2024年第2期40卷

作者： Shin, Yejin Choi, Yujin Won, Jaeseung Hong, Taehoon Koo, Choongwan Incheon Natl Univ Div Architecture & Urban Design Incheon 22012 South Korea Sogang Univ Dept Artificial Intelligence Seoul 04107 South Korea Yonsei Univ Dept Architecture & Architectural Engn Seoul 03722 South Korea

The integration of computer vision technology into construction sites poses various challenges due to the complex environment. Prior studies on computer vision related to heavy construction equipment has primarily focused on a limited range of equipment types provided in standard databases, such as the Microsoft Common objects in Context (MS COCO) data set. The conventional approach has limitations in capturing the diverse working conditions and dynamic environments encountered in real construction sites. To overcome the challenge, this study proposes a new benchmark model for the automated detection and classification of a wide range of heavy construction equipment (i.e., nine representative types) commonly used in construction sites by using a deep convolution neural network. This study was conducted in four steps: (1) data collection and preparation, (2) data transformation, (3) model training, and (4) model validation. The proposed you only look once (YOLO)v5l (large, YOLOv5 with a larger network) model demonstrated high reliability, achieving a mean average precision (mAP)_0.5 : 0.95 of 90.26%. This study makes a significant contribution to the domain of construction engineering and management by providing a more efficient and systematic management system to proactively prevent heavy equipment-related safety accidents with diverse working conditions and dynamic environments encountered at construction sites. Moreover, the proposed approach can be extended to integrate advanced techniques such as case-based reasoning, digital twin, and blockchain, allowing for the automated activity recognition in various occlusions, the carbon emissions monitoring and diagnostics of heavy equipment, and a robust real-time construction management system with enhanced security.

关键词： Construction site Heavy equipment Benchmark model object detection and classification Computer vision Field applicability

来源：评论

学校读者我要写书评

暂无评论

iSense: Intelligent object Sensing and Robot Tracking Through Networked Coupled Magnetic Resonant Coils

引用

IEEE INTERNET OF THINGS JOURNAL 2021年第8期8卷 6637-6648页

作者： Li, Kai Muncuk, Ufuk Naderi, M. Yousof Chowdhury, Kaushik R. Northeastern Univ Dept Elect & Comp Engn Boston MA 02120 USA

object sensing and tracking using electric and magnetic fields allow intelligent interaction, automation, and adaptation in cyber-physical systems. Our approach, called iSense, uses a software-defined collaborative sensing technique for the detection of the type of object when placed on a large surface and tracking its mobility. iSense is cost effective, low power, and scalable, which allows its use over large surfaces. First, we introduce a dual-coil magnetic resonant sensing architecture based on nested coils, i.e., passive (outer) and active (inner) coils, for low-power contactless sensing. Second, a data-driven support vector machine-based approach helps to classify different types of objects using the voltage readings obtained at the passive coil. iSense combines sensed voltage information from multiple different coils spread over the surface with a group-based interference mitigation mechanism between coils for collaborative sensing. We validate our system with real-time prototype and experimental evaluations. We demonstrate the detection of seven different types of objects over three different materials, and real-time detection and tracking of mobile objects including a robot car. Experimental results show that each sensing coil only consumes few milliwatts, i.e.,18x less than inductive sensing and 15x less than classical magnetic resonance sensing, extend sensing depth to 3 cm, and enable tracking on the large surface sensing with more than 90% accuracy for velocity estimation.

关键词： Contactless sensing magnetic resonance mobile car tracking object detection and classification

来源：评论

学校读者我要写书评

暂无评论

Multiple object detection and classification on Uneven Terrain using Multi-Channel LIDAR for UGV

Multiple Object Detection and Classification on Uneven Terra...

引用

Conference on Unmanned Systems Technology XIV

作者： Cho, Kuk Baeg, Seung-Ho Park, Sangdeok Univ Sci & Technol 176 Gajung Dong Deajeon 305350 South Korea KITECH Ansan 426791 Gyeonggi Do South Korea

ISBN: (纸本)9780819490650

In this paper, we proposed multiple object detection and classification methods using a multi-channel light detection and ranging (LIDAR) device for an unmanned ground vehicle (UGV) in dense forest. The natural terrain environment is described by trees and bushes which are representative wild objects. To describe them on a navigation map, we must detect and classify objects with a single LIDAR sensor rather than multiple sensors on an UGV. It has the advantage of price, communication, and synchronization. If scanning object data are similar with real data, it is better recognizable;a large amount of 3D point cloud data can resemble object in appearance. There is a trade-off relationship between point cloud data and acquisition time. Getting a number of items of point cloud data gradually increases with measurement time and computation time. Our approach for achieving multiple object detection and classification consists of three steps that start from the raw 3D point cloud data. The 3D point cloud data are composed of one frame by one scan. First, the point cloud data are divided by two groups, ground points and non-ground points. Second, the non-ground points are separated by vegetation points and object points. Finally, the object points are arranged as a set of feasible objects with a clustering method. In our approach, random sample consensus (RANSAC), support vector machine (SVM) with principle component analysis (PCA) and fuzzy c-mean (FCM) algorithms belong to each step to derive multiple objects detection and classification. In a local statistic approach, the density of a point cloud in the processing area influences the system performance. Our primary research goal is to achieve improved performance with a data set of one frame per scan for an UGV. Thus, we evaluate the proposed object detection and classification algorithm with a single scan frame which is acquired from three different LIDAR devices. The device has different scan channels. The 1-channel an

关键词： 3D LIDAR object detection and classification UGV uneven terrain

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：