检索结果-内蒙古大学图书馆

deep learning-based autofocus method enhances image quality in light-sheet fluorescence microscopy

BIOMEDICAL OPTICS EXPRESS 2021年第8期12卷 5214-5226页

作者： Li, Chen Moatti, Adele Zhang, Xuying Ghashghaei, H. Troy Greenabum, Alon North Carolina State Univ Joint Dept Biomed Engn Raleigh NC 27695 USA Univ North Carolina Chapel Hill Raleigh NC 27695 USA North Carolina State Univ Comparat Med Inst Raleigh NC 27695 USA North Carolina State Univ Dept Mol Biomed Sci Raleigh NC 27695 USA North Carolina State Univ Bioinformat Res Ctr Raleigh NC 27695 USA

Light-sheet fluorescence microscopy (LSFM) is a minimally invasive and high throughput imaging technique ideal for capturing large volumes of tissue with sub-cellular resolution. A fundamental requirement for LSFM is a seamless overlap of the light-sheet that excites a selective plane in the specimen, with the focal plane of the objective lens. However, spatial heterogeneity in the refractive index of the specimen often results in violation of this requirement when imaging deep in the tissue. To address this issue, autofocus methods are commonly used to refocus the focal plane of the objective-lens on the light-sheet. Yet, autofocus techniques are slow since they require capturing a stack of images and tend to fail in the presence of spherical aberrations that dominate volume imaging. To address these issues, we present a deep learning-based autofocus framework that can estimate the position of the objective-lens focal plane relative to the light-sheet, based on two defocused images. This approach outperforms or provides comparable results with the best traditional autofocus method on small and large image patches respectively. When the trained network is integrated with a custom-built LSFM, a certainty measure is used to further refine the network's prediction. The network performance is demonstrated in real-time on cleared genetically labeled mouse forebrain and pig cochleae samples. Our study provides a framework that could improve light-sheet microscopy and its application toward imaging large 3D specimens with high spatial resolution. (c) 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

关键词： High throughput optics image metrics image quality Spatial resolution Spherical aberrations Three dimensional imaging

来源：评论

学校读者我要写书评

暂无评论

Automatic segmentation for synchrotron-based imaging of porous bread dough using deep learning approach

引用

JOURNAL OF SYNCHROTRON RADIATION 2021年第2期28卷 566-575页

作者： Ali, Salah Mayo, Sherry Gostar, Amirali K. Tennakoon, Ruwan Bab-Hadiashar, Alireza MCann, Thu Tuhumury, Helen Favaro, Jenny RMIT Univ Sch Engn Melbourne Vic Australia CSIRO Mfg Clayton Vic Australia RMIT Univ Sch Sci Melbourne Vic Australia CSIRO Agr & Food Werribee Vic Australia

In recent years, major capability improvements at synchrotron beamlines have given researchers the ability to capture more complex structures at a higher resolution within a very short time. This opens up the possibility of studying dynamic processes and observing resulting structural changes over time. However, such studies can create a huge quantity of 3D image data, which presents a major challenge for segmentation and analysis. Here tomography experiments at the Australian synchrotron source are examined, which were used to study bread dough formulations during rising and baking, resulting in over 460 individual 3D datasets. The current pipeline for segmentation and analysis involves semi-automated methods using commercial software that require a large amount of user input. This paper focuses on exploring machine learning methods to automate this process. The main challenge to be faced is in generating adequate training datasets to train the machine learning model. Creating training data by manually segmenting real images is very labour-intensive, so instead methods of automatically creating synthetic training datasets which have the same attributes of the original images have been tested. The generated synthetic images are used to train a U-Net model, which is then used to segment the original bread dough images. The trained U-Net outperformed the previously used segmentation techniques while taking less manual effort. This automated model for data segmentation would alleviate the time-consuming aspects of experimental workflow and would open the door to perform 4D characterization experiments with smaller time steps.

关键词： micro-CT deep learning bread automatic analysis micro-structure porosity

来源：评论

学校读者我要写书评

暂无评论

Oral Cancer Using deep learning and Auto-Fluorescence image Analysis

Oral Cancer Using Deep Learning and Auto-Fluorescence Image ...

引用

2024 International Conference on Advances in Modern Age Technologies for Health and Engineering Science, AMATHE 2024

作者： Muhammed Yaseer, P. Arul Xavier, V.M. Shyni, S.S. Karunya Institute Of Technology And Sciences Division Of Computer Science And Engineering Tamilnadu Coimbatore India Karunya Institute Of Technology And Sciences Division Of Data Science And Cyber Security Tamilnadu Coimbatore India

ISBN: (数字)9798350371567

ISBN: (纸本)9798350371567

Poor treatment outcomes result from the fact that oral cancer is frequently identified at an advanced stage, making it a serious and sometimes lethal disease. Thus, it is essential to develop quick and accurate procedures for detecting oral cancer in its early stages. In this work, we present a unique non-invasive detection technique for evaluating auto-fluorescence signals in oral tissues utilizing convolutional neural networks ResNet50 architecture. Biological tissues naturally produce auto-fluorescence signals, and variations in these signals may be a marker of the existence of malignant cells. Our approach seeks to precisely identify malignant and normal oral tissues by capturing these minute variations in auto-fluorescence patterns. Using a sizable dataset of pictures of oral tissue, we trained and evaluated our RestNet50 model, and the results showed impressive accuracy rates of 89%. The better performance of our detection method can be attributed to the effective capture of high-level features made possible by the ResNet50 architecture. Furthermore, the suggested approach demonstrated quick processing speeds, which allowed it to be used for real-time clinical applications. This approach may greatly enhance patient prognosis and treatment results by identifying oral cancer in its early stages with high accuracy of 89% when compared to VGG16 and VGG19. Although more validation research is necessary before the suggested approach can be used in clinical settings, it shows promise in facilitating early diagnosis and intervention to enhance patient outcomes. © 2024 IEEE.

关键词： Fluorescence

来源：评论

学校读者我要写书评

暂无评论

FPGA-Based deep Convolutional Neural Network Optimization Method

FPGA-Based Deep Convolutional Neural Network Optimization Me...

引用

2021 International Conference on Signal processing and Machine learning, CONF-SPML 2021

作者： Wen, Lilan Southwest Jiaotong University SWJTU-Leeds Joint School Chengdu China

ISBN: (纸本)9781665417341

With the increasing demand for computing speed and real-time data processing in various fields, deep learning and convolutional neural networks are more and more widely used in the field of computer vision. FPGA-based deep convolutional neural networks (CNN) have been proposed and developed rapidly due to its high parallel processing ability, portability, and low power consumption. To further improve the network efficiency, this paper studies the software acceleration tool Vivado HLS provided by Xilinx, the quantification and pruning of convolution neural network model, which can effectively optimize the network model and accelerate the reasoning process. © 2021 IEEE.

关键词： High level synthesis

来源：评论

学校读者我要写书评

暂无评论

deep Convolutional Neural Networks for Rail Surface Defect Perception

Deep Convolutional Neural Networks for Rail Surface Defect P...

引用

image processing, Computer Vision and Machine learning (ICICML), International Conference on

作者： Yinjie Wang Ting Wang Zihao Yang Shuoqi Ren Xiaohui Gouliu Jialei Gao School of Mechanical Engineering Lanzhou Jiaotong University Lanzhou China School of Mathematics and Physics Lanzhou Jiaotong University Lanzhou China School of Materials Science and Engineering Lanzhou Jiaotong University Lanzhou China

Rail surface defect inspection is a vital element for ensuring railway safety. We critically analyze the application of deep convolutional neural networks (CNN) in Rail Surface Defect Perception (RSDP). We scrutinize 43 studies, revealing how CNN, YOLO, R-CNN, and Semantic Segmentation techniques, have revolutionized RSDP by surpassing traditional inspection methods in speed, accuracy, and efficiency. Our examination underscores the necessity for comprehensive and standardized datasets to support the varied and complex nature of rail surface defects. Additionally, we highlight the persistent challenges such as limited defect samples and the imperative for real-time detection capabilities. The paper details the performance and adaptation of various CNN models for RSDP, charting a course for future inquiry. We advocate for enhanced datasets, unified defect classification, and refined deep learning models, aiming to bolster the progress in CNN-driven RSDP technology for enhanced railway safety.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multi-modal Conditional Bounding Box Regression for Music Score Following 29

Multi-modal Conditional Bounding Box Regression for Music Sc...

引用

29th European Signal processing Conference (EUSIPCO)

作者： Henkel, Florian Widmer, Gerhard Johannes Kepler Univ Linz Inst Computat Percept Linz Austria Johannes Kepler Univ Linz LIT Artificial Intelligence Lab Linz Austria

ISBN: (纸本)9789082797060

This paper addresses the problem of sheet-image-based on-line audio-to-score alignment also known as score following. Drawing inspiration from object detection, a conditional neural network architecture is proposed that directly predicts x,y coordinates of the matching positions in a complete score sheet image at each point in time for a given musical performance. Experiments are conducted on a synthetic polyphonic piano benchmark dataset and the new method is compared to several existing approaches from the literature for sheet-image-based score following as well as an Optical Music Recognition baseline. The proposed approach achieves new state-of-the-art results and furthermore significantly improves the alignment performance on a set of real-world piano recordings by applying Impulse Responses as a data augmentation technique.

关键词： audio-to-score alignment score following conditional object detection multi-modal deep learning

来源：评论

学校读者我要写书评

暂无评论

tsegGAN: A Generative Adversarial Network for Segmenting Touching Nontext Components From Text Ones in Handwriting

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2021年 70卷 1-10页

作者： Mondal, Riktim Bhowmik, Showmik Sarkar, Ram Jadavpur Univ Dept Comp Sci & Engn Kolkata 700032 India Ghani Khan Choudhury Inst Engn & Technol Dept Comp Sci & Engn Malda 732141 India

Segmentation of a touching component to separate its constituent text and nontext parts is always a very crucial but challenging task toward developing a comprehensive document image processing (DIP) system. This is because, irrespective of document types, either printed or handwritten, the nontext parts need to be suppressed first before processing the text parts through an optical character recognition (OCR) system. Although a good number of attempts have been made to address this issue for printed documents, the same for regular handwritten document images is almost none. However, the appearance of touching components where a nontext part gets joined with a text part is a common issue in freestyle handwriting. To this end, in this work, we tailor-make a generative adversarial network (GAN)-based model with a suitable loss function that we name tsegGAN. We also prepare an in-house data set by collecting touching components from different real-world handwritten documents to evaluate our model. The performance comparison of our model with state-of-the-art GAN models shows that tsegGAN has outperformed the others with a significant margin.

关键词： deep learning document image processing (DIP) generative adversarial network (GAN) handwriting image segmentation optical character recognition (OCR) text nontext separation touching component

来源：评论

学校读者我要写书评

暂无评论

Implementation and Optimization of Target Detection based on Multi-core DSP 10

Implementation and Optimization of Target Detection based on...

引用

10th IEEE Joint International Information Technology and Artificial Intelligence Conference, ITAIC 2022

作者： Yang, Tao Jia, Qingzhong Tao, Ximing Huang, Hong Beijing Institute of Technology Key Laboratory of Dynamic and Control of Flight Vehicle Ministry of Education Beijing China

ISBN: (数字)9781665422079

ISBN: (纸本)9781665422079

As an important branch in the field of image processing, target detection requires more and more real-time algorithm with its wide application in industry, agriculture, medical and military industries, which also puts forward higher requirement for the optimization and application of embedded platform. There are two kinds of target detection algorithms: traditional detection algorithm and deep-learning-based detection algorithm. In the embedded platform, due to the limitation of hardware performance, the detection algorithm based on deep learning is difficult to realize because of high complexity. Therefore, the improvement, optimization and transplantation of traditional algorithms have become hot spots. In this paper, a detection algorithm based on Canny edge feature matching is proposed, and multi-core parallel development of TMS320C6678 platform is carried out. Parallelism of the algorithm is improved by using methods such as image chunking, vectorized data packaging, optimization of data flow, etc. By connecting the video board based on FPGA for physical verification, the speed of the algorithm after multi-core optimization is increased by 7.6 times, which has certain engineering practical significance. © 2022 IEEE.

关键词： Digital signal processing

来源：评论

学校读者我要写书评

暂无评论

Pavement image Data Set for deep learning: A Synthetic Approach

Pavement Image Data Set for Deep Learning: A Synthetic Appro...

引用

International Airfield and Highway Pavements Conference of the Transportation-and-Development-Institute (T and DI) of the American-Society-of-Civil-Engineers (ASCE)

作者： Gong, Haitao Wang, Feng Texas State Univ Ingram Sch Engn San Marcos TX 78666 USA

ISBN: (数字)9780784483503

ISBN: (纸本)9780784483503

deep learning methods have shown a promising approach to reliable automated pavement condition survey in recent years. However, the training of models requires large quantities of annotated data, which is normally time consuming, expensive, and sometimes difficult to obtain. This research aims to explore the viability of using synthetic pavement image data to train convolutional neural networks (CNNs) for automated pavement crack detection. A procedural approach of generating synthetic pavement crack image data is proposed. Perlin noise is adopted to mimic the real-world cracks, and simple textures are used to control the generated crack type. Mask R-CNN is used to train on the synthetic data developed in this study. Both synthetic and real data sets are used to evaluate the performance of the trained model. The results indicate that training a crack detection model using only synthetic data can reach almost the same level of accuracy as using the real data.

关键词： Pavements

来源：评论

学校读者我要写书评

暂无评论

A Literature Review of Machine learning Techniques for Dance Recognition and Robotic Vision

A Literature Review of Machine Learning Techniques for Dance...

引用

Robotics and Technologies for Industrial Automation (ROBOTHIA), IEEE International Conference on

作者： Lee Wei San Samuel-Soma M. Ajibade Muhammed Basheer Jasser Adefemi Ayodele Babatunde Adedotun Ajayi Mbiatke Anthony Bassey Department of Data Science and Artificial Intelligence Faculty of Engineering and Technology Sunway University Selangor Darul Ehsan Malaysia Research Centre for Nanomaterials and Energy Technology (RCNMET) Faculty of Engineering and Technology Sunway University Selangor Malaysia Research Centre for Human-Machine Collaboration (HUMAC) Faculty of Engineering and Technology Sunway University Petaling Jaya Selangor Malaysia University of East London London UK Department of Cyber security Faculty of Computing & Informatics Ladoke Akintola University of Technology Ogbomoso Nigeria Department of Business Administration Universiti Tun Hussein Onn Malaysia Batu Pahat Malaysia

ISBN: (数字)9798350356755

ISBN: (纸本)9798350356762

image recognition, powered by machine learning (ML), has significantly advanced applications in both dance movement recognition and robotic vision. This review examines key ML techniques, including Convolutional Neural Networks (CNNs), deep Neural Networks (DNNs), Self-Organizing Maps (SOMs), and Long Short-Term Memory (LSTM) networks, alongside pose estimation methods like OpenPose and Part Affinity Fields (PAFs). These techniques enhance dance classification, real-time feedback, and motion analysis, with OpenPose + LSTMs and PAFs + LSTMs demonstrating the highest accuracy. Notwithstanding progress, obstacles such as high computational costs, data dependency, and real-time implementation challenges persist. Beyond dance, these methods are critical in robotic vision, intelligent automation, and industrial image processing, enabling autonomous robotic navigation, defect detection in manufacturing, and AI-driven motion tracking. By leveraging human movement analysis for robotics, ML improves human-robot interaction, robotic-assisted rehabilitation, and industrial automation. Despite progress, challenges such as high computational demands, data dependency, and real-time constraints remain. This review explores future directions, including multimodal data fusion, hybrid AI models, and real-time optimization, bridging the gap between AI-driven motion systems and intelligent automation to enhance adaptability and efficiency across domains.

关键词： Intelligent automation Humanities Solid modeling image recognition Accuracy Service robots Robot sensing systems real-time systems Robots Long short term memory

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：