检索结果-内蒙古大学图书馆

YOLO Based Crack Detection of Structures with Edge Detection Process

Transactions of the Korean Institute of Electrical Engineers 2024年第12期73卷 2391-2397页

作者： Kim, Ji-Soo Chung, Kyungyong Division of AI Computer Science and Engineering Kyonggi University Korea Republic of

In this paper, we propose an algorithm that combines the YOLO model with edge detection techniques to effectively detect cracks occurring in structures. Existing manual inspection methods are time-consuming, costly, and suffer from poor accuracy. To solve this problem, this study proposes an automated deep learning-based detection method, specifically using a YOLO model and a canny edge detection algorithm to perform more precise crack boundary extraction. The proposed algorithm consists of two steps. The first step is image preprocessing, which uses several preprocessing techniques such as gray-scale conversion, median blur, normalization, and brightness inversion to highlight the edges of cracks. In the second step, the canny edge detection algorithm is used to detect the boundaries of the cracks in the preprocessed image, and the YOLOv8 model is used to learn and detect the location of the corresponding cracks. The results show that the best performance is achieved when the YOLO model is trained using 6 steps of preprocessing and canny edge detection. Copyright © The Korean Institute of Electrical Engineers.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Research on Memory-Efficient Approach to Sleep Quality Estimation for Edge Devices 15

Research on Memory-Efficient Approach to Sleep Quality Estim...

引用

15th International Conference on Information and Communication Technology Convergence, ICTC 2024

作者： Kang, Seungeun Kim, Jin Kim, Namgi Kyonggi University Division of Ai Computer Science and Engineering Suwon Korea Republic of

ISBN: (纸本)9798350364637

Lifelog data has become easier to collect due to the development of wearable technology and IoT devices, and its importance has increased with the growth of the healthcare industry. In particular, personalized services such as personalized healthcare and stress management are possible because of the personal characteristics of lifelog data. In this study, we build a sleep quality classification system using lifelog datasets. As a preprocessing of the time series data, we propose a method for using different time periods depending on the type of question to be considered, a method for synchronizing variable lengths, and a process for normalizing the data. In addition, this paper suggests Weighted Validation Loss Propagation and Set Lower Bound methods to enhance generalization performance in the process of building and training classification deep learning models with Convolution Neural Network (CNN) and Attention-based models. Then, we discuss model significance, applications, and limitations. We also propose future research to overcome the limitations. © 2024 IEEE.

关键词： Sleep research

来源：评论

学校读者我要写书评

暂无评论

Explainable Anomaly Detection Using Vision Transformer Based SVDD

引用

computers, Materials & Continua 2023年第3期74卷 6573-6586页

作者： Ji-Won Baek Kyungyong Chung Department of Computer Science Kyonggi UniversitySuwon-si 16227Korea Division of AI Computer Science and Engineering Kyonggi UniversitySuwon-si 16227Korea

Explainable ai extracts a variety of patterns of data in the learning process and draws hidden information through the discovery of semantic *** is possible to offer the explainable basis of decision-making for inference *** the causality of risk factors that have an ambiguous association in big medical data,it is possible to increase transparency and reliability of explainable decision-making that helps to diagnose disease *** addition,the technique makes it possible to accurately predict disease risk for anomaly *** transformer for anomaly detection from image data makes classification through ***,in MLP,a vector value depends on patch sequence information,and thus a weight *** should solve the problem that there is a difference in the result value according to the change in the *** addition,since the deep learning model is a black box model,there is a problem that it is difficult to interpret the results determined by the ***,there is a need for an explainablemethod for the part where the disease *** solve the problem,this study proposes explainable anomaly detection using vision transformerbasedDeep Support Vector Data Description(SVDD).The proposed method applies the SVDD to solve the problem of MLP in which a result value is different depending on a weight change that is influenced by patch sequence information used in the vision *** order to draw the explainability of model results,it visualizes normal parts through *** health data,both medical staff and patients are able to identify abnormal parts *** addition,it is possible to improve the reliability of models and medical *** performance evaluation normal/abnormal classification accuracy and f-measure are evaluated,according to whether to apply *** Results The results of classification by applying the proposed SVDD are evaluated ***,through the proposed meth

关键词： Explainable ai anomaly detection vision transformer SVDD health care deep learning classification

来源：评论

学校读者我要写书评

暂无评论

Dynamic Framerate SlowFast Network for Improving Autonomous Driving Performance

引用

IEIE Transactions on Smart Processing and Computing 2023年第3期12卷 261-268页

作者： Jeon, Byeong-Uk Chung, Kyungyong Department of Computer Science Kyonggi University Suwon Korea Republic of Division of AI Computer Science and Engineering Kyonggi University Suwon Korea Republic of

computer vision technology is used for autonomous driving and road traffic safety. Accordingly, studies on deep learning models that detect and analyze objects through images or videos are ongoing. On the other hand, deep learning algorithms that detect even the action of things require high computing performance. In addition, the computing performance of autonomous driving vehicles for processing such tasks is inferior. These classification processes are not used for recognizing and determining autonomous driving vehicles because it is impossible to process the classification of the action of autonomous driving vehicles in an autonomous driving vehicle on a real-time basis. This paper proposes a Dynamic Framerate SlowFast network for improving autonomous driving performance. Unlike pre-existing studies, the proposed model includes a cropping process through the YOLO model. In addition, it measures the similarity between unit frames through the SSIM and skips the input when the similarity exceeds a certain level. This process made it possible to reduce the number of frames entered into the model. Compared to the existing SlowFast Network, the performance evaluation compared the time required to analyze one image and the AUC of classification results when the number of input frames was reduced through similarity analysis techniques. The similarity analysis technique achieved the highest AUC when the SSIM was applied. The Dynamic Framerate SlowFast network proposed in this study achieved an AUC of 0.7126 and took an FPS of 0.7912 to analyze the entire verification video data. Compared to the pre-existing SlowFast network, which took an FPS of 0.5285 to achieve an AUC of 0.7531, the Dynamic Framerate SlowFast network achieved faster and more accurate results. Therefore, using the proposed technique, it is possible to achieve faster detection results while maintaining the object action detection AUC of the SlowFast network. © 2023 Institute of Electronics and Informatio

关键词： Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

YOLO based Object Features Detection using Non-Local Means Denoising and Data Augmentation

引用

Transactions of the Korean Institute of Electrical Engineers 2022年第9期71卷 1280-1285页

作者： Park, Geon Kang, Ye-Yeon Kim, Gyu-Il Chung, Kyungyong Division of AI Computer Science and Engineering Kyonggi University Korea Republic of

When fruits are harvested in farms, most of them go through a manual sorting process and classify and distribute decomposed fruits. However, there is a limit to manually classifying large amounts in a situation where the number of workers is decreasing in farms. To solve this problem, it is important to divide normal and decomposed fruits in real time to minimize the proportion of manpower used in the screening process. We propose a method of YOLO based Object Features Detection Using Non-Local Means Denoising and Data Augmentation. The proposed method collects desired data through the Crawling method, and the preprocessing minimizes image noise through Non-Local Means Denoising. The built image dataset uses YOLOv3, an object detection algorithm, to distinguish and detect normal and decayed fruits. As a result of performance evaluation, object detection of YOLOv3 objects in a proposed method rather than detection results shows that Recall increases 10% performance and increases 9% in the remaining Recall and IoU. Therefore, the proposed method can increase screening efficiency by detecting decayed fruits well. © 2022 Korean Institute of Electrical Engineers. All rights reserved.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Multiple-Object Tracking Using Histogram Stamp Extraction in CCTV Environments

引用

computers, Materials & Continua 2023年第12期77卷 3619-3635页

作者： Ye-Yeon Kang Geon Park Hyun Yoo Kyungyong Chung Division of AI Computer Science and Engineering Kyonggi UniversitySuwon16227Korea Contents Convergence Software Research Institute Kyonggi UniversitySuwon16227Korea

Object tracking,an important technology in the field of image processing and computer vision,is used to continuously track a specific object or person in an *** technology may be effective in identifying the same person within one image,but it has limitations in handling multiple images owing to the difficulty in identifying whether the object appearing in other images is the *** tracking the same object using two or more images,there must be a way to determine that objects existing in different images are the same ***,this paper attempts to determine the same object present in different images using color information among the unique information of the ***,this study proposes a multiple-object-tracking method using histogram stamp extraction in closed-circuit television *** proposed method determines the presence or absence of a target object in an image by comparing the similarity between the image containing the target object and other *** this end,a unique color value of the target object is extracted based on its color distribution in the image using three methods:mean,mode,and interquartile *** Top-N accuracy method is used to analyze the accuracy of each method,and the results show that the mean method had an accuracy of 93.5%(Top-2).Furthermore,the positive prediction value experimental results show that the accuracy of the mean method was 65.7%.As a result of the analysis,it is possible to detect and track the same object present in different images using the unique color of the *** the results,it is possible to track the same object that can minimize manpower without using personal information when detecting objects in different *** the last response speed experiment,it was shown that when the mean was used,the color extraction of the object was possible in real time with 0.016954 *** this,it is possible to detect and track the same object in real time when using the proposed

关键词： Data mining deep learning object detection object tracking real-time object detection multiple object image processing

来源：评论

学校读者我要写书评

暂无评论

Research on Building K-Contents Specialized Text2Image Pipeline 27

Research on Building K-Contents Specialized Text2Image Pipel...

引用

27th International Conference on Advanced Communications Technology, ICACT 2025

作者： Kang, Seungeun Lee, Dahyun Han, Sangbum Lim, Seokjae Kim, Namgi Division of AI Computer Science and Engineering Kyonggi University Suwon154-42 Korea Republic of AI Tech Team LOTTE INNOVATE Seoul179 Korea Republic of

ISBN: (纸本)9791188428137

Recently, there has been a lot of research on the Text2Image generative model as social and technological interest in generative models has increased. In addition, the interest in Korean contents (K-Contnets) has increased. In this paper, we propose "K-Contents Specialized Text2Image Pipeline". "K-Contents Specialized Text2Image Pipelilne" is specialized for understanding Korean prompts and aims to generate images specialized for various K-Contents. To build the pipeline, we collect and preprocess text and image data. Then, we train the Korean Text Encoder on the text data and train three specialized K-Contents generation pipelines on the image data. We also propose to use the SDXL model’s Img2Img Refiner with the SD model for time-efficient image generation. After that, we use the learned pipelines to generate real images and analyze their performance. Finally, we discuss the utilization of the above pipelines, limitations, and future research. Copyright 2025 Global IT Research Institute (GIRI). All rights reserved.

关键词： Image Generation K-Contents Korean text encoder Stable Diffusion

来源：评论

学校读者我要写书评

暂无评论

Enhancing Radiology Report Interpretation with Large-Scale Language Models: A Two-Stage Fine-Tuning Approach 27

Enhancing Radiology Report Interpretation with Large-Scale L...

引用

27th International Conference on Advanced Communications Technology, ICACT 2025

作者： Pi, Sun-Woo Park, Myeong-Soo Lee, Byoung-Dai Department of Computer Science Graduate School Kyonggi University Suwon Korea Republic of Division of AI and Computer Engineering Kyonggi University Suwon Korea Republic of

ISBN: (纸本)9791188428137

Radiology reports contain complex medical terminology and specialized knowledge, making them difficult for both patients and medical professionals to interpret. This study aims to address this challenge by developing a large-scale language model specifically designed for interpreting chest radiology reports. We focus on four key natural language processing (NLP) tasks—summarization, paraphrasing, abbreviation interpretation, and question answering—using a synthetic dataset derived from the MIMIC-CXR reports and GPT-3.5 Turbo. To enhance the model’s performance, we propose a two-stage supervised fine-tuning (SFT) process, incorporating real-world medical data from PubMedQA and MedQA, in addition to the synthetic dataset. The resulting models, Model-1 and Model-2, were evaluated based on accuracy, conciseness, and clarity, using test data not seen during training. Experimental results demonstrated that the proposed two-stage SFT method achieved strong performance across all four tasks, providing comparable performance to models such as GPT-3.5, Bard, Llama2, and MedAlpaca in key evaluation metrics, despite using a relatively smaller number of parameters. These findings suggest that synthetic data, when combined with domain-specific datasets, can significantly improve the interpretive capabilities of large-scale language models in the medical domain. Copyright 2025 Global IT Research Institute (GIRI). All rights reserved.

关键词： Radiology

来源：评论

学校读者我要写书评

暂无评论

Augmented and End-to-End Models for Defect Classification of Structures 18

Augmented and End-to-End Models for Defect Classification of...

引用

18th ACM International Conference on Distributed and Event-Based Systems, DEBS 2024

作者： Kim, Gyu-Il Chung, Kyungyong Department of Computer Science Kyonggi University Suwon Korea Republic of Division of AI Computer Science and Engineering Kyonggi University Suwon Korea Republic of

ISBN: (纸本)9798400704437

This paper proposes an automated defect classification model for the safety diagnosis and maintenance of concrete structures. Traditional manual classification methods are time-consuming and rely on subjective judgment, whereas the proposed model utilizes time-series data obtained from sensors to enable faster and more accurate defect detection through data augmentation and Transformer-based classification. Data augmentation compensates for the imbalance in defect data distribution, and the Transformer effectively interprets the complex temporal patterns of the data, capturing long-range dependencies. This end-to-end model performs an integrated process from data processing to defect classification. Future research will explore the model's generalizability and the potential for integration into real-time monitoring systems. This paper presents a new approach to improving the economic and efficient evaluation of the safety of concrete structures. © 2024 Owner/Author.

关键词： Defects

来源：评论

学校读者我要写书评

暂无评论

Search and Recommendation Systems with Metadata Extensions 26

Search and Recommendation Systems with Metadata Extensions

引用

26th International Conference on Advanced Communications Technology, ICACT 2024

作者： Kim, Woo-Hyeon Kim, Joo-Chang Kyonggi University Division of Ai Computer Science and Computer Engineering Korea Republic of Contents Convergence Software Research Institute Kyonggi University Korea Republic of

ISBN: (纸本)9791188428120

This paper proposes an ai-based video metadata extension model to overcome the limitations of video search and recommendation systems in the multimedia industry. Current video searches and recommendations utilize pre-added metadata. Metadata includes filenames, keywords, tags, genres, etc. This makes it impossible to make direct predictions about the content of a video without pre-added metadata. These platforms also analyze your previous search history, viewing history, etc. to understand your interests in order to serve you personalized videos. This may not reflect the actual content and may raise privacy concerns. In addition, recommendation systems suffer from a cold start problem, which is the lack of an initial target, as well as a bubble effect. Therefore, this study proposes a search and recommendation system by expanding metadata in videos using techniques such as shot boundary detection, speech recognition, and text mining. The proposed method selects the main objects required by the recommendation system based on the object frequency and extracts the corresponding objects from the video frame by frame. In addition, we extract the speech from the video separately, convert the speech to text to extract the script and apply text mining techniques to the extracted script to quantify it. Then, we synchronize the object frequency and the transcript to create a single contextual data. After that, we group videos and clips based on the contextual data and index them. Finally, we utilize Shot Boundary Detection to segment videos based on their content. To ensure that the generated contextual data is appropriate for the video, the proposed model compares the extracted script with the video's subtitle data to check and calibrate its accuracy. The model can then be fine-tuned by tuning and cross-validating the hyperparameter to improve its performance. These models can be incorporated into a variety of content discovery and recommendation platforms. By using expanded

关键词： Speech recognition

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：