检索结果-内蒙古大学图书馆

A simple deep learning based image illumination correction method for paintings

PATTERN RECOGNITION LETTERS 2020年 138卷 392-396页

作者： Goswami, Suranjan Singh, Satish Kumar IIIT Allahabad IT Allahabad 211015 Uttar Pradesh India

image illumination correction has been a long standing topic for research in the Computer Vision problem. However, all previous literature on this topic has either been statistical in nature in the sense that a specified algorithm has been developed for approaching a particular case of illumination normalization, or involves extremely complex deep learning methods for illumination correction of either one of over illuminated or under illuminated images. We present here a very simple deep learning based image illumination correction architecture which works on color images of paintings irrespective of whether they are under or over illuminated. We have tested the results using a synthetic database as well as on real world painting images of diverse nature. (C) 2020 Elsevier B.V. All rights reserved.

关键词： deep learning algorithm Low space-time complexity image enhancement RGB color image High range variation

来源：评论

学校读者我要写书评

暂无评论

Multi-scale, multi-dimensional binocular endoscopic image depth estimation network

引用

COMPUTERS IN BIOLOGY AND MEDICINE 2023年第1期164卷 107305-107305页

作者： Wang, Xiongzhi Nie, Yunfeng Ren, Wenqi Wei, Min Zhang, Jingang Univ Chinese Acad Sci Sch Future Technol Beijing 100039 Peoples R China Xidian Univ Sch Aerosp Science&Technol Xian 710071 Peoples R China Vrije Univ Brussel & Flanders Make Dept Appl Phys & Photon Brussel Photon B-1050 Brussels Belgium Chinese Acad Sci State Key Lab Informat Secur Inst Informat Engn Beijing 100093 Peoples R China Chinese Peoples Liberat Army Gen Hosp Med Ctr 4 Dept Orthoped Beijing 100853 Peoples R China

During invasive surgery, the use of deep learning techniques to acquire depth information from lesion sites in real-time is hindered by the lack of endoscopic environmental datasets. This work aims to develop a high-accuracy three-dimensional (3D) simulation model for generating image datasets and acquiring depth information in real-time. Here, we proposed an end-to-end multi-scale supervisory depth estimation network (MMDENet) model for the depth estimation of pairs of binocular images. The proposed MMDENet highlights a multi-scale feature extraction module incorporating contextual information to enhance the correspondence precision of poorly exposed regions. A multi-dimensional information-guidance refinement module is also proposed to refine the initial coarse disparity map. Statistical experimentation demonstrated a 3.14% reduction in endpoint error compared to state-of-the-art methods. With a processing time of approximately 30fps, satisfying the requirements of real-time operation applications. In order to validate the performance of the trained MMDENet in actual endoscopic images, we conduct both qualitative and quantitative analysis with 93.38% high precision, which holds great promise for applications in surgical navigation.

关键词： Depth estimation Endoscopic datasets Convolutional neural network Stereoscopic vision

来源：评论

学校读者我要写书评

暂无评论

Ball and Goal image Recognition on Humanoid Robot Darwin OP Using Faster Region-Based Convolutional Neural Networks (Faster R-CNN) Method 22

Ball and Goal Image Recognition on Humanoid Robot Darwin OP ...

引用

Proceedings of the 7th International Conference on Sustainable Information Engineering and Technology

作者： Randy Christian Saputra Mochammad Zava Abbiyansyah Fitri Utaminingrum Brawijaya University Indonesia)

ISBN: (纸本)9781450397117

In humanoid robot soccer, the capacity to precisely track a ball is a crucial problem that is made challenging by processing limits and the subsequent inability to interpret all data from a high-definition image. This research suggests a method for locating and sizing balls in a computationally effective field setting. This research presents an enhanced, Faster Region-Based CNN-based deep learning architecture for multi-class ball and goal recognition. The proposed framework incorporates improved Faster RCNN model development, data argumentation, ball and goal image library building, and performance assessment. This study is a pioneer in employing 1000 real-world photographs to build a multi-labeled image class ball and goal. The convolutional and pooling layers are also improved for more precise and quick identification. The test findings reveal that the suggested method outperformed conventional detectors regarding detecting accuracy and processing speed. It has excellent potential for use in developing an autonomous, real-time image recognition system for humanoid robots.

关键词： Humanoid Robots Faster Region-Based CNN image Recognition

来源：评论

学校读者我要写书评

暂无评论

An Intelligent Evaluation Method of Root Canal Therapy Quality Based on deep learning

An Intelligent Evaluation Method of Root Canal Therapy Quali...

引用

2022 Chinese Automation Congress, CAC 2022

作者： Liu, Jie Peng, Gang Yan, Shiqian Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Hubei Eya Medical Investment Management Co. Ltd Wuhan430061 China

ISBN: (纸本)9781665465335

Dentists judge the quality of root canal therapy for each patient very time-consuming, and inefficient, lack of quantitative evaluation criteria, easy to cause judgment errors. At the same time, the traditional method of extracting the root canal image features based on experience is difficult to accurately extract the filling area of the root canal and the tooth area where it is located, resulting in low accuracy of root canal target segmentation, and thus affecting the accuracy of root canal treatment quality evaluation. In this paper, for the real root canal treatment images of patients, the CenterNet network was used to detect the root canal target of root canal images. Then, according to the test results, the U-Net full convolution neural network was used to segment the root canal filling region and its tooth region. Finally, the segmented images were quantitatively evaluated according to the professional evaluation index of doctors. The experimental results show that the coincidence rate between the evaluation results of this method and the evaluation results of professional doctors is 83.3 %, which can better complete the quality evaluation of root canal therapy for patients, effectively improve the work efficiency of doctors, and has reference significance for the application of artificial intelligence in medical field. © 2022 IEEE.

关键词： Patient treatment

来源：评论

学校读者我要写书评

暂无评论

Recognition of grape leaf diseases using MobileNetV3 and deep transfer learning

引用

International Journal of Agricultural and Biological Engineering 2022年第3期15卷 184-194页

作者： Xiang Yin Wenhua Li Zhen Li Lili Yi School of Agricultural Engineering and Food Science Shandong University of TechnologyZibo 255000ShandongChina School of Artificial Intelligence Nanjing University of Aeronautics and AstronauticsNanjing 210016China

timely diagnosis and accurate identification of grape leaf diseases are decisive for controlling the spread of disease and ensuring the healthy development of the grape *** objective of this research was to propose a simple and efficient approach to improve grape leaf disease identification accuracy with limited computing resources and scale of training image dataset based on deep transfer learning and an improved MobileNetV3 model(GLD-DTL).A pre-training model was obtained by training MobileNetV3 using the imageNet dataset to extract common features of the grape *** the last convolution layer of the pre-training model was modified by adding a batch normalization function.A dropout layer followed by a fully connected layer was used to improve the generalization ability of the pre-training model and realize a weight matrix to quantify the scores of six diseases,according to which the Softmax method was added as the top layer of the modified networks to give probability distribution of six ***,the grape leaf diseases dataset,which was constructed by processing the image with data augmentation and image annotation technologies,was input into the modified networks to retrain the networks to obtain the grape leaf diseases recognition(GLDR)*** showed that the proposed GLD-DTL approach had better performance than some recent *** identification accuracy was as high as 99.84%while the model size was as small as 30 MB.

关键词： grape leaf diseases real-time recognition deep transfer learning MobileNetV3

来源：评论

学校读者我要写书评

暂无评论

A Brief Overview of deep learning based Techniques for the Detection of Wheat Leaf Disease: A Recent Study

A Brief Overview of Deep Learning based Techniques for the D...

引用

International Conference on ent Computing and Control Systems (ICICCS)

作者： Protyush Protim Neog Salil Batra Sudhir Saraswat Emani Likith Sharma P. Pavan Kumar Ankit Kumar Pandey Department of Computer Science and Engineering Lovely Professional University Phagwara Punjab

Wheat is an important cereal crop and is the second most consumed cereal after rice globally. It is a staple food for more than one-third of the world’s population. The production of wheat depends on various factors, such as climate, temperature, soil, pests, bacteria, and other biotic and abiotic factors. However, diseases can have a significant impact on wheat production. Various wheat diseases can affect crop yields,including leaf rust, leaf spot, spike infection, virus, bacterial back chaff, bacterial spike blight, and aphids. Leaf rust, in particular, is known to cause severe damage to wheat leaves. To combat these diseases, researchers have been investigating the use of advanced technologies such as deep learning and image-processing approaches for plant disease recognition. The process of disease detection involves several steps, including image preprocessing, segmentation, feature extraction, and classification. The accuracy of these steps directly affects the reliability and accuracy of the classification algorithms used to identify plant diseases. Recent studies have reviewed the state-of-the-art techniques used in this context and evaluated their effectiveness in real life applications. Overall, the use of advanced technologies such as deep learning and image processing for disease detection in wheat crops holds great promise in improving crop yields and reducing losses due to diseases. By identifying diseases at an early stage, farmers can take appropriate measures to control the spread of the disease and prevent significant crop damage. This research work presents a comparative analysis of the state-of-the-art work that has been carried out in this context and the effectiveness of the techniques in real-time.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A SAR Target Recognition Strategy Guided by Electromagnetic Scattering Feature

A SAR Target Recognition Strategy Guided by Electromagnetic ...

引用

Signal, Information and Data processing (ICSIDP), IEEE International Conference on

作者： Yifei Yin Liang Chen Lujiao Liu Yufan Meng Fan Chen Hao Shi Beijing Institute of Technology National Key Laboratory of Science and Technology on Space-Born Intelligent Information Processing Beijing China Aerospace Information Research Institute Chinese Academy of Sciences Beijing China

ISBN: (数字)9798331515669

ISBN: (纸本)9798331515676

SAR target Automatic Target Recognition (ATR) is indispensable in SAR image interpretation. Recently, deep learning technology has been widely used in SAR target recognition tasks. Most networks achieve incremental improvements in target recognition by modifying their structures to extract visual features of targets. However, due to the unique imaging mechanism, relying solely on visual features often leads to the loss of target information. In contrast, the ASC model, which captures the electromagnetic scattering characteristics of the target, plays a crucial role in target recognition tasks. Unfortunately, traditional parameter estimation methods for extracting the ASC model are computationally expensive and time-consuming, making them impractical for real-world applications. To address these issues, we propose a novel target recognition method based on electromagnetic scattering features in this paper. First, a lightweight network-based feature extraction module is designed. Then, the target ASC image is used as the ground truth for guidance, with image intensity and target structure serving as the loss functions during training. Finally, an ASC model-guided feature fusion network is designed, utilizing the fused features for target recognition. On the MSTAR dataset, a visual assessment experiment demonstrated that the proposed feature extraction module effectively extracts electromagnetic scattering features under various operating conditions. Subsequently, in downstream classification tasks, the inclusion of the proposed module resulted in improved accuracy compared to other networks. Additionally, a visualization analysis of the classification network showed that, under the guidance of electromagnetic scattering features, the network achieved good interpretability.

关键词： Training deep learning Visualization Parameter estimation Target recognition Computational modeling Electromagnetic scattering Imaging Feature extraction Radar polarimetry

来源：评论

学校读者我要写书评

暂无评论

Visual Object Tracking Via Multi-Stream deep Similarity learning Networks

引用

IEEE TRANSACTIONS ON image processing 2020年 29卷 3311-3320页

作者： Li, Kunpeng Kong, Yu Fu, Yun Northeastern Univ Dept Elect & Comp Engn Coll Engn Boston MA 02115 USA Rochester Inst Technol B Thomas Golisano Coll Comp & Informat Sci Rochester NY 14623 USA Northeastern Univ Khoury Coll Comp Sci Boston MA 02115 USA

Visual tracking remains a challenging research problem because of appearance variations of the object over time, changing cluttered background and requirement for real-time speed. In this paper, we investigate the problem of real-time accurate tracking in a instance-level tracking-by-verification mechanism. We propose a multi-stream deep similarity learning network to learn a similarity comparison model purely off-line. Our loss function encourages the distance between a positive patch and the background patches to be larger than that between the positive patch and the target template. Then, the learned model is directly used to determine the patch in each frame that is most distinctive to the background context and similar to the target template. Within the learned feature space, even if the distance between positive patches becomes large caused by the interference of background clutter, impact from hard distractors from the same class or the appearance change of the target, our method can still distinguish the target robustly using the relative distance. Besides, we also propose a complete framework considering the recovery from failures and the template updating to further improve the tracking performance without taking too much computing resource. Experiments on visual tracking benchmarks show the effectiveness of the proposed tracker when comparing with several recent real-time-speed trackers as well as trackers already included in the benchmarks.

关键词： deep learning visual tracking

来源：评论

学校读者我要写书评

暂无评论

Efficient Context Integration through Factorized Pyramidal learning for Ultra-Lightweight Semantic Segmentation

arXiv

引用

arXiv 2023年

作者： Atif, Nadeem Mazhar, Saquib Sarma, Debajit Bhuyan, M.K. Ahamed, Shaik Rafi Dept. of Electronics and Electrical Engineerning IIT Guwahati Guwahati781039 India

Semantic segmentation is a pixel-level prediction task to classify each pixel of the input image. deep learning models, such as convolutional neural networks (CNNs), have been extremely successful in achieving excellent performances in this domain. However, mobile application, such as autonomous driving, demand real-time processing of incoming stream of images. Hence, achieving efficient architectures along with enhanced accuracy is of paramount importance. Since, accuracy and model size of CNNs are intrinsically contentious in nature, the challenge is to achieve a decent trade-off between accuracy and model size. To address this, we propose a novel Factorized Pyramidal learning (FPL) module to aggregate rich contextual information in an efficient manner. On one hand, it uses a bank of convolutional filters with multiple dilation rates which leads to multi-scale context aggregation;crucial in achieving better accuracy. On the other hand, parameters are reduced by a careful factorization of the employed filters;crucial in achieving lightweight models. Moreover, we decompose the spatial pyramid into two stages which enables a simple and efficient feature fusion within the module to solve the notorious checkerboard effect. We also design a dedicated Feature-image Reinforcement (FIR) unit to carry out the fusion operation of shallow and deep features with the downsampled versions of the input image. This gives an accuracy enhancement without increasing model parameters. Based on the FPL module and FIR unit, we propose an ultra-lightweight real-time network, called FPLNet, which achieves state-of-the-art accuracy-efficiency trade-off. More specifically, with only less than 0.5 million parameters, the proposed network achieves 66.93% and 66.28% mIoU on Cityscapes validation and test set, respectively. Moreover, FPLNet has a processing speed of 95.5 frames per second (FPS). Copyright © 2023, The Authors. All rights reserved.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

real-time based Violence Detection from CCTV Camera using Machine learning Method

Real-time based Violence Detection from CCTV Camera using Ma...

引用

2022 International Conference on Industry 4.0 Technology, I4Tech 2022

作者： Silva Deena, J. Ahammed, Md. Tabil Boppana, Udaya Mouni Afroj, Maharin Ghosh, Sudipto Hossain, Sohaima Balaji, Priyadharshini St. Joseph's Institute of Technology Semmancheri Department of Ece Tamil Nadu Chennai India Khulna University of Engineering and Technology Department of Ece Khulna Bangladesh Universiti Tun Hussein Onn Malaysia Johor Malaysia Bangladesh University of Business and Technology Department of Cse Dhaka Bangladesh Jeppiaar Engineering College Department of Electrical and Electronics Engineering Tamil Nadu Chennai India

ISBN: (数字)9781665471961

ISBN: (纸本)9781665471961

Based on deep-learning approaches, we developed a real-time violence detector for surveillance video systems. In the model given here (overall generality-accuracy-fast response time), CNN serves as a space feature extractor, while LSTM is used to learn time-related relationships. Due to the large number of devices that can record video from camera systems, like those used in surveillance systems, body-worn webcams, and phones, it has become hard to keep track of video footage from many surveillance devices. Using crowd-based video footage, we analyzes and alerts possible those who are affected by violent material is found in the clip. Keeping an eye on huge crowds during social gatherings, especially those where there is a possibility of violence becomes very difficult. The speed, accuracy, and generality of violent event detectors across a variety of video sources and formats are all factors that go into determining their usefulness. Intelligent monitoring technology has been extensively deployed in the nation in recent years to continually support the development of a safe city. Behavioral intelligence analysis is becoming more popular in the realm of intelligent image analysis. Currently, complex activities such as fighting or violence are rarely studied in behavior analysis techniques;instead, they focus on basic movements such as running or leaping. To preserve social order and safeguard people's lives and property, competent and intelligent analysis of violence through video surveillance is vital. To that end, we've put up an overview of the most recent techniques for spotting violent scenes in recorded video. © 2022 IEEE.

关键词： Video signal processing

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：