检索结果-内蒙古大学图书馆

8th International Scientific conference on Intelligent Information Technologies for Industry, IITI 2024

作者： Kuang, Di Guo, Zhichang Song, Fanghui School of Mathematics Harbin Institute of Technology Heilongjiang Harbin150001 China

ISBN: (纸本)9783031774102

image segmentation is an important and complex task in computer vision. The variational level set method has become a popular approach for image segmentation due to its topological invariance. However, the evolution process may be unstable, leading to segmentation failures due to excessively flat or sharp final evolutionary profiles. In order to improve the accuracy and stability of evolution, we propose a high-order level set variational model, which applies the variational form of the molecular beam epitaxy (MBE) equation. The model avoids the need for re-initialization. In order to calculate the model, we drive the gradient flow of the model. We propose a finite difference semi-implicit and semi-explicit scheme. Additionally, we also design the scalar auxiliary variable scheme of the model, which expands its application, and can be solved by fast algorithms. Numerical experiments show that the model is outperforms current mainstream models in the processing of noisy images and medical images, and can be applied to actual image segmentation problems. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Molecular beam epitaxy

来源：评论

学校读者我要写书评

暂无评论

Techniques for Pre-processing CT Angiography (CTA) images for the Diagnosis of Cardiac Disease 2

Techniques for Pre-processing CT Angiography (CTA) Images fo...

引用

2nd International conference on Augmented Intelligence and Sustainable systems, ICAISS 2023

作者： Punitha, T. Santhi Preethika, S K. Piramu Vistas School of Computing Sciences Chennai India

ISBN: (纸本)9798350325799

Artificial Intelligence (AI) is a rapidly developing discipline that concentrates on teaching computers to comprehend and scrutinize images, particularly in the medical field, with a specific emphasis on diagnosing Cardiovascular Diseases. Coronary Computed Tomography Angiography (CCTA) is a widely accepted non-invasive diagnostic examination used to evaluate cardiac disease (CD). CCTA images are coronary artery anatomy, whereas functional stress testing assesses for inducible cardiac ischemia. Nevertheless, CCTA images frequently encounter diverse forms of noise and artifacts that can deteriorate their quality and hinder precise diagnosis. Preprocessing CT Angiography images is a vital undertaking in medical diagnosis utilizing computer vision. This study commences by employing image enhancement preprocessing methods to diminish noise, blurring, and image weakening. The initial stages involve utilizing edge detection algorithms and smoothing/filtering functions to enhance pixel clarity and edge definition in CCTA images using Python programming language and the PyCharm Integrated Development Environment (IDE) are implemented in this work. © 2023 IEEE.

关键词： image enhancement

来源：评论

学校读者我要写书评

暂无评论

Invisible Backdoor Attack with image Contours Triggers 24th

Invisible Backdoor Attack with Image Contours Triggers

引用

24th International conference on algorithms and Architectures for Parallel processing, ICA3PP 2024

作者： Hou, Yangzhi Yang, Ying Jinan China Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing Shandong Fundamental Research Center for Computer Science Jinan China

ISBN: (纸本)9789819615445

In recent years, backdoor attack techniques on neural networks have been widely studied and researched. In this attack mode, the model implanted with a backdoor behaves normally when processing normal inputs, but once it encounters a special trigger designed by the attacker, its prediction results will be maliciously manipulated. Different backdoor attack strategies have been proposed to improve the stealth and robustness of the triggers. However, most of the existing backdoor attack strategies use fixed triggering patterns or input-independent triggering patterns, and thus are easily detected and blocked by existing defense systems. To address these limitations, we propose a backdoor attack method based on image contours. Specifically, we first use image contours as the target region for backdoor injection and fill them with specific color information to generate trigger patterns. Then, using our designed Generative Network (GN) and the Auxiliary Extraction Network (AE), the trigger patterns are covertly embedded into the images to be contaminated to generate a poisoned dataset, which contaminates the training model for the purpose of malicious attacks. Compared with traditional methods, our trigger patterns are more natural and harder to detect. Experimental results show that our proposed attack method achieves excellent performance in terms of both stealth and robustness. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Classification of Aerial images and Applications: A Deep Learning Perspective 16

Classification of Aerial Images and Applications: A Deep Lea...

引用

16th International conference on Computational Intelligence and Communication Networks

作者： Monika, Pralayakaveri Reddy, Chinta Midhilesh Venkata Sai Ramesh, S. SRM Inst Sci & Technol Sch Comp Coll Engn & Technol Dept Comp Technol Chennai Tamil Nadu India

ISBN: (纸本)9798331505264;9798331505271

Aerial image detection is a crucial technology for a variety of applications, including but not limited to urban planning, environmental monitoring, and disaster management. In this project, we implement deep learning algorithms, in particular the AlexNet model to classify aerial images taken from different landscapes. The dataset includes cityscape, road, and nature images, which are resized and normalized to make sure these image inputs conform to the input size for the AlexNet model. Fine-tuning means changing the top layer of AlexNet to have the corresponding number of classes as that of the dataset. Using pre-trained weights helps speed up the time it will take to learn the model set up in pytorch. We use momentum Stochastic Gradient Descent (SGD) to optimize the model, and a learning rate scheduler to boost the performance. The data is then divided into training and testing sets, and the accuracy of the model is assessed using the test set. The classification results indicate that fine-tuned alexnet model work well in different aerial image classification. We present the full potential of deep learning for fast aerial image classifications, which will serve as an accurate tool for aerial image classification with applications in disaster response, urban development and environmental monitoring. This model can be further improved for different datasets and for real time image processing problems.

关键词： Aerial image detection Deep learning AlexNet Transfer learning Fine-tuning PyTorch Stochastic gradient descent (SGD)

来源：评论

学校读者我要写书评

暂无评论

Detection of Indian Fake Currency using image processing

Detection of Indian Fake Currency using Image Processing

引用

2022 International conference on Augmented Intelligence and Sustainable systems, ICAISS 2022

作者： Asha Banu, S.M. Sandhya, S. Sundari, T vijaya Shri Ranjani, P.R. Dept of ECE Sri Krishna College of Engineering and Technology India

ISBN: (纸本)9781665489621

The fake currency notes are detected using image processing employing MATLAB in this paper. This project aims in providing the best techniques in image acquisition, and image segmentation. The work uses CANNY's algorithm to extract the notes' features much more effectively. algorithms for image processing are used to extract the options. The technique used here functions well with the recently introduced denominations of 500 and 2000. The recommended strategy provides a practical means of detecting fake currency that is supported by physical inspection. © 2022 IEEE.

关键词： Fake detection

来源：评论

学校读者我要写书评

暂无评论

Gigastep - One Billion Steps per Second Multi-agent Reinforcement Learning 37

Gigastep - One Billion Steps per Second Multi-agent Reinforc...

引用

37th conference on Neural Information processing systems (NeurIPS)

作者： Lechner, Mathias Yin, Lianhao Seyde, Tim Wang, Tsun-Hsuan Xiao, Wei Hasani, Ramin Rountree, Joshua Rus, Daniela MIT CSAIL Cambridge MA 02139 USA US Air Force Washington DC 20330 USA

ISBN: (纸本)9781713899921

Multi-agent reinforcement learning (MARL) research is faced with a trade-off: it either uses complex environments requiring large compute resources, which makes it inaccessible to researchers with limited resources, or relies on simpler dynamics for faster execution, which makes the transferability of the results to more realistic tasks challenging. Motivated by these challenges, we present Gigastep, a fully vectorizable, MARL environment implemented in JAX, capable of executing up to one billion environment steps per second on consumer-grade hardware. Its design allows for comprehensive MARL experimentation, including a complex, high-dimensional space defined by 3D dynamics, stochasticity, and partial observations. Gigastep supports both collaborative and adversarial tasks, continuous and discrete action spaces, and provides RGB image and feature vector observations, allowing the evaluation of a wide range of MARL algorithms. We validate Gigastep's usability through an extensive set of experiments, underscoring its role in widening participation and promoting inclusivity in the MARL research community. MIT licensed code is available at https://***/mlech26l/gigastep.

关键词： Multi agent systems

来源：评论

学校读者我要写书评

暂无评论

An application of smart visual air conditioning controller to enhance control efficiency in office environments

引用

RESULTS IN ENGINEERING 2025年 25卷

作者： Tian, Yu Zhao, Zilong Loo, Junn Yong Univ Malaya Fac Engn Dept Mech Engn Kuala Lumpur 50603 Malaysia Monash Univ Malaysia Sch Informat Technol Jalan Lagoon Selatan 47500 Bandar Sunway Selangor Darul Malaysia

Innovative solutions for sustainability and energy efficiency are crucial in green building management. This study presents a novel approach to optimizing air conditioning (AC) system operations in commercial buildings, with a focus on real-time control aimed at reducing energy consumption. We propose the Smart visual Air Conditioning Controller (SVACC), which utilizes computer vision and deep learning-based human detection to intelligently manage AC operation, minimizing unnecessary runtime. By detecting human presence in meeting rooms, the system dynamically adjusts AC activation based on occupancy, thereby significantly reducing energy waste. A statistical analysis conducted over five months across ten conference rooms demonstrated that the SVACC reduced AC usage time by 33.60 %. We validate and optimize the SVACC across various building types, including commercial office spaces, industrial warehouses and laboratories, and residential apartments. The system achieved an optimal balance with 96.55 % precision and 93.33 % recall, resulting in an F1 score of 0.9492, demonstrating high performance across various environments. Our results underscore the effectiveness of the SVACC, which highlights the potential of integrating advanced deep learning models with HVAC systems to optimize energy consumption. This approach offers a promising solution for improving HVAC design and energy management across diverse building environments. Future work will focus on refining sensor technology and control algorithms to further optimize energy efficiency.

关键词： HVAC Building management Deep learning image processing Computer vision

来源：评论

学校读者我要写书评

暂无评论

Survey on Deep Learning based Automated systems for the Detection and Grading of Diabetic Retinopathy using Retinal Fundus images

Survey on Deep Learning based Automated Systems for the Dete...

引用

2023 International conference on Power, Instrumentation, Control and Computing, PICC 2023

作者： vijayan, vinya Salim, A. College of Engineering Trivandrum Affiliated to Apj Abdul Kalam Technological University Dept. of Computer Science and Engineering Kerala India Government Engineering College Thrissur Affiliated to Apj Abdul Kalam Technological University Dept. of Computer Science and Engineering Kerala India

ISBN: (纸本)9798350334463

Diabetic Retinopathy is a common cause of blindness in people diagnosed as diabetic and it is developed as a subsequent condition by diabetes mellitus. Retinal fundus image analysis is commonly used for the computer aided diagnosis of diabetic retinopathy and its grading. Different automated systems based on deep learning techniques are available for the detection and grading of diabetic retinopathy using fundus image analysis. This article summarizes and analyses the performance of different deep learning based diabetic retinopathy detection and grading methods. Different types of image processing techniques used by the deep learning based algorithms and performance indicators used in the classification process are also discussed. This article also listed the publicly available datasets for Diabetic retinopathy detection and discussed the challenges that further need to be addressed in this area. © 2023 IEEE.

关键词： Grading

来源：评论

学校读者我要写书评

暂无评论

Fish Detection and Quantity Estimation Based on Sonar images 1st

Fish Detection and Quantity Estimation Based on Sonar Images

引用

1st International conference on Extended Reality, ICXR 2024

作者： Du, Wenxiang Yan, Shuai Qi, Yue State Key Laboratory of Virtual Reality Technology and Systems Beihang University Beijing100191 China Qingdao Research Institute of Beihang University Qingdao266100 China

ISBN: (纸本)9789819636785

Due to the complex underwater environment and dense targets, it is difficult to quickly and accurately calculate the location and quantity of fish targets in marine ranching. We have developed an underwater target detection and counting algorithm based on sonar images, which includes four parts: image preprocessing, background subtraction, contour detection, and target counting. Firstly, we design a sliding-window-based gain algorithm based on the uneven grayscale value of sonar images, which amplifies the effective signal while smoothing the grayscale of the image. Secondly, background subtraction is used to separate the foreground and background of the sonar image, which can remove background noise and filter the target using a filtering algorithm to generate a binary image with clear targets. Then, target detection is combined with morphological processing techniques and contour detection algorithms. Finally, we use image erosion technology to separate overlapping targets, calculate the contour position of the targets, and count the quantity. A large number of results indicate that our algorithm can quickly and accurately detect the position of fish in sonar images while also counting the number of fishes. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Binary images

来源：评论

学校读者我要写书评

暂无评论

Fake Face image Classification by Blending the Scalable Convolution Network and Hierarchical vision Transformer 1

引用

4th International conference on Computer and Communication Technologies, IC3T 2022

作者： Kerenalli, Sudarshana Yendapalli, Vamsidhar Mylarareddy, C. Department of CSE SoT GITAM University Karnataka Bengaluru562163 India

ISBN: (数字)9789811985638

ISBN: (纸本)9789811985621

A face has been used as a primary and unique attribute to authenticate individual users in emerging security approaches. Cybercriminals use the double-edged sword "image processing" capabilities to deceive innocent users. The underlying technology is based on advanced machine learning and deep learning algorithms. The intentions of cyber criminals range from simple mimicking or trolling to creating violent situations in society. Hence, it is necessary to resolve such problems by identifying the fake face images generated by expert humans or artificial intelligent algorithms. Machine learning and artificial neural networks are used to resolve the issue. In this work, we have designed an approach for detecting deep learning-generated fake face images by combining the capabilities of the scalable convolutional neural networks (CNN) "EfficientNet" and hierarchical vision transformer (viT) "shifted window transformer". The proposed method accurately classifies the fake face images with a 98.04% accuracy and a validation loss of 0.1656 on the 140 k_real_fake_faces image dataset. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Generative adversarial networks

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：