检索结果-内蒙古大学图书馆

A Comprehensive Survey of Image Augmentation techniques for Deep Learning

PATTERN RECOGNITION 2023年第1期137卷

作者： Xu, Mingle Yoon, Sook Fuentes, Alvaro Park, Dong Sun Jeonbuk Natl Univ Dept Elect Engn Jeonbuk 54896 South Korea Mokpo Natl Univ Dept Comp Engn Jeonnam 58554 South Korea Jeonbuk Natl Univ Core Res Inst Intelligent Robots Jeonbuk 54896 South Korea

Although deep learning has achieved satisfactory performance in computer vision, a large volume of im-ages is required. However, collecting images is often expensive and challenging. Many image augmenta-tion algorithms have been proposed to alleviate this issue. Understanding existing algorithms is, therefore, essential for finding suitable and developing novel methods for a given task. In this study, we perform a comprehensive survey of image augmentation for deep learning using a novel informative taxonomy. To examine the basic objective of image augmentation, we introduce challenges in computer vision tasks and vicinity distribution. The algorithms are then classified among three categories: model-free, model-based, and optimizing policy-based. The model-free category employs the methods from image process-ing, whereas the model-based approach leverages image generation models to synthesize images. In con-trast, the optimizing policy-based approach aims to find an optimal combination of operations. Based on this analysis, we believe that our survey enhances the understanding necessary for choosing suitable methods and designing novel algorithms.(c) 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( http://***/licenses/by-nc-nd/4.0/ )

关键词： Image augmentation Deep learning Image variation Vicinity distribution Data augmentation computer vision

来源：评论

学校读者我要写书评

暂无评论

Autonomous mental development in high dimensional state and action spaces

Autonomous mental development in high dimensional state and ...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： A. Joshi Juyang Weng Department of Electrical and Computer Engineering Michigan State University East Lansing MI USA Department of Computer Science and Engineering Michigan State University East Lansing MI USA

Autonomous mental development (AMD) of robots opened a new paradigm for developing machine intelligence, using neural network type of techniques and it fundamentally changed the way an intelligent machine is developed from manual to autonomous. The work presented is a part of SAIL (self-organizing autonomous incremental learner) project which deals with autonomous development of entire humanoid robot with vision, audition, manipulation and locomotion. The major issue addressed is the challenge of high dimensional action space (5 to 10) in addition to the high dimensional context state space (hundreds to thousands and beyond), typically required by an AMD machine. This is the first work that studies a high dimensional (numeric) action space in conjunction with a high dimensional perception (context state) space, under the AMID mode. Two new learning algorithms, Direct Update on Direction Cosines (DUDC) and High-Dimensional Conjugate Gradient Search (HCGS), are developed, implemented and tested. The convergence properties of both the algorithms and their targeted applications are discussed. Autonomous learning of speech production under reinforcement learning is studied as an example.

关键词： Autonomous mental development intelligent robots Orbital robotics Machine intelligence Neural networks intelligent networks Learning systems Manuals Humanoid robots State-space methods

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Survey of Image Augmentation techniques for Deep Learning

arXiv

引用

arXiv 2022年

作者： Xu, Mingle Yoon, Sook Fuentes, Alvaro Park, Dong Sun Department of Electronics Engineering Jeonbuk National University Jeonbuk54896 Korea Republic of Department of Computer Engineering Mokpo National University Jeonnam58554 Korea Republic of Core Research Institute of Intelligent Robots Jeonbuk National University Jeonbuk54896 Korea Republic of

Although deep learning has achieved satisfactory performance in computer vision, a large volume of images is required. However, collecting images is often expensive and challenging. Many image augmentation algorithms have been proposed to alleviate this issue. Understanding existing algorithms is, therefore, essential for finding suitable and developing novel methods for a given task. In this study, we perform a comprehensive survey of image augmentation for deep learning using a novel informative taxonomy. To examine the basic objective of image augmentation, we introduce challenges in computer vision tasks and vicinity distribution. The algorithms are then classified among three categories: model-free, model-based, and optimizing policy-based. The model-free category employs the methods from image processing, whereas the model-based approach leverages image generation models to synthesize images. In contrast, the optimizing policy-based approach aims to find an optimal combination of operations. Based on this analysis, we believe that our survey enhances the understanding necessary for choosing suitable methods and designing novel algorithms. © 2022, CC BY-NC-ND.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

Generic object recognition: building coarse 3D descriptions from line drawings

Generic object recognition: building coarse 3D descriptions ...

引用

Workshop on Interpretation of 3D Scenes

作者： R. Bergevin M.D. Levine Computer Vision and Robotics Laboratory McGill Research Center for Intelligent Machines McGill University Montreal QUE Canada

PARVO, a computer vision system that addresses the problems of fast and generic recognition of unexpected 3D objects from single 2D views, is introduced. Recently, RBC (recognition by components), a new human image understanding theory, has been proposed on the basis of the results of various psychological studies. However, no systematic computational evaluation of its many aspects has been reported yet. The object recognition system the authors have built is a first step toward this goal, since its design respects and makes explicit the main assumptions of the proposed theory. It analyzes single-view 2D line drawings of 3D objects typical of the ones used in human image understanding studies. The main issues related to generic object recognition are discussed, original algorithms and techniques specific to the author's systems are described, and results of the different processing stages of the system are presented.< >

关键词： Object recognition computer vision Humans Psychology intelligent robots Robot vision systems Laboratories intelligent structures Machine intelligence Visual system

来源：评论

学校读者我要写书评

暂无评论

vision-based lane detection for an autonomous ground vehicle: A comparative field test

Vision-based lane detection for an autonomous ground vehicle...

引用

Southeastern Symposium on System Theory

作者： Forrest N. Bush Joel M. Esposito US Naval Academy Annapolis MD USA

ISBN: (纸本)9781424456901

We examine the problem of designing computer vision algorithms to autonomously drive an off road vehicle between two lane markings painted on the ground. In this paper we describe field tests used to compare the efficacy of two popular line extractions techniques from the literature: the Hough Transform and the RANSAC Algorithm. Although it is very implementation dependent, we found the Hough Transform to be superior to the RANSAC algorithm in both speed and accuracy for identifying lane markings in the off road environment.

关键词： Vehicle detection Land vehicles Testing Road vehicles Remotely operated vehicles computer vision Mobile robots Global Positioning System Road transportation intelligent vehicles

来源：评论

学校读者我要写书评

暂无评论

3D feature based mapping towards mobile robots' enhanced performance in rescue missions

3D feature based mapping towards mobile robots' enhanced per...

引用

2009 IEEE/RSJ International Conference on intelligent robots and Systems

作者： Paloma de la Puente Diego Rodriguez-Losada Alberto Valero Fernando Matia Intelligent Control Group Universidad Politecnica de Madrid Madrid Spain

This paper presents a feature based 3D mapping approach with regard to obtaining compact models of semi-structured environments such as partially destroyed buildings where mobile robots are to carry out rescue activities. To gather the 3D data, we use a laser scanner, employing a nodding data acquisition system mounted on both real and simulated robots. Our segmentation algorithm comes up from the integration of computer vision techniques, allowing for a fast separation of points corresponding to different, not necessarily planar, surfaces. The subsequent extraction of geometrical features out of each region's points is done by means of least-squares fitting. A Maximum Incremental Probability algorithm formulated upon the Extended Kalman Filter provides 6D localization and produces a map of planar patches with a convex-hull based representation. Scenarios from the Unified System for Automation and Robot Simulation (USARSim), including world models from past RoboCup Rescue editions' arenas, have been utilized to conduct some of our experiments.

关键词： Mobile robots Robotics and automation intelligent robots Laser modes Feature extraction Solid modeling Iterative algorithms Iterative closest point algorithm Simultaneous localization and mapping Noise measurement

来源：评论

学校读者我要写书评

暂无评论

Defect Detection in Porcelain Industry Based on Deep Learning techniques

Defect Detection in Porcelain Industry Based on Deep Learnin...

引用

International Symposium on Symbolic and Numeric algorithms for Scientific Computing (SYNASC)

作者： Adriana Birlutiu Adrian Burlacu Manuella Kadar Daniela Onita “Decembrie 1918” University of Alba Iulia Alba Iulia Romania “Gheorghe Asachi” Technical University of Iasi Iasi Romania

ISBN: (纸本)9781538626276

This paper presents an automated defect management system based on machine learning and computer vision that detects and quantifies different types of defects in porcelain products. The system is developed in collaboration with an industrial porcelain producer and integrates robots, artificial vision and machine learning. At present, in most of the companies involved in the porcelain industry, defect detection is performed manually by employees. An intelligent system for product monitoring and defect detection is very much needed. Our proposed system is implemented through a convolutional neural network which analyzes images of the products and predicts if the product is defective or not. Experimental evaluation on an image data set acquired at the industrial partner shows promising results. The proposed architecture will finally have a positive economic impact for the company by optimizing the production flow and reducing the production costs.

关键词： Feature extraction Inspection Porcelain Production Industries computer vision

来源：评论

学校读者我要写书评

暂无评论

Learning-based control of preception for mobility

Learning-based control of preception for mobility

引用

IEEE International Symposium on intelligent Control (ISIC)

作者： M. Barth S. Das B. Bhanu College of Engineering University of California Riverside CA USA

To overcome the lack of flexibility and inadequacy in performance speed of perception systems for use in real-time tasks, the authors have applied integrated learning techniques to a perception system that is based on a selective sensing paradigm. The incorporation of multiple learning algorithms at different levels provides a great deal of flexibility and robustness when different perceptual task are performed. Using a selective sensing paradigm allows the system to eliminate a large amount of nonpertinent sensory data so that processing speed is greatly increased. Such a perception system is being implemented on an autonomous mobile agent. The methodology and a preliminary example of learning within the perception system are presented.< >

关键词： Mobile agents Navigation Application software Mobile robots Learning systems computer vision Machine learning Genetic algorithms Parallel processing Hardware

来源：评论

学校读者我要写书评

暂无评论

Reexamining Indoor Mobile robots from a Cognitive Perspective

Reexamining Indoor Mobile Robots from a Cognitive Perspectiv...

引用

Robotics and computer vision (ICRCV), International Conference on

作者： Fei Song Jinyu Li Fengzhen Tang Yandong Tang Bailu Si State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China Chinese Academy of Sciences Institutes for Robotics and Intelligent Manufacturing Shenyang China University of Chinese Academy of Sciences Beijing China School of Systems Science Beijing Normal University Beijing China

The field of traditional mobile robot navigation has undergone a gradual transformation, evolving into a standardized and procedural research domain. Through a fresh cognitive perspective on this navigation process, a bridge emerges between robotic researchers and cognitive neuroscientists, thereby fostering the growth of interdisciplinary exploration. This article meticulously examines three pivotal phases of robot navigation: information acquisition, simultaneous localization and mapping (SLAM), and path planning. These phases are intricately linked to pertinent cognitive research. Furthermore, a comprehensive survey of existing biomimetic and neuromorphic algorithms is presented, highlighting their applications in augmenting robot navigation capabilities. Ultimately, this article charts a visionary course for the future of robot navigation, envisioning its potential in facilitating more intelligent and efficient navigation techniques.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Development of computer vision Based Aluminium Scrap Sorting Using Real Time Metal Scrap Dataset

Development of Computer Vision Based Aluminium Scrap Sorting...

引用

Sustainable Expert Systems (ICSES), International Conference on

作者： Adhithya S J. Kathirvelan School of Electronics Engineering Vellore Institute of Technology Vellore India

ISBN: (数字)9798331540364

ISBN: (纸本)9798331540371

The aluminium manufacturing industries face numerous challenges when it comes to sorting metal scraps, particularly when scraps are small in size. Traditional methods like flotation and gravity separation are inefficient, energy-intensive, and often fail when dealing with size variations. Recent advancements in computer vision and deep learning algorithms, including intelligent robotic systems, offer promising solutions. However, challenges such as accurate detection in complex environments, varying lighting conditions, and overlapping objects still persist. Newer techniques like YOLO (You Only Look Once) for real-time object detection, and Transfer Learning to improve model accuracy with smaller datasets, have shown potential. This research aims to overcome these challenges by developing an AI-powered, computer vision-based sorting system that can identify and classify aluminium scraps by color and shape at high speed. The authors developed a custom dataset and trained models using CNN architectures such as DenseNet-161, ResNet-152, and VGG 19. By employing ensemble techniques, this research achieved an 75% of accuracy.

关键词： Deep learning YOLO computer vision Accuracy Service robots Computational modeling Aluminum Real-time systems Recycling Sorting

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：