• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.343-345
• /
• 2024

얼굴 연령 분류 기법은 신원 확인 시스템 고도화, 유동 인구 통계 자동화 시스템 구축, 연령 제한 콘텐츠 관리 시스템 고도화 등 다양한 분야에 적용할 수 있는 확장 가능성을 가진다. 넓은 확장 가능성을 가지는 만큼 적용된 시스템의 안정성을 위해서는 얼굴 연령 분류 기법의 높은 정확도는 필수적이다. 따라서, 본 논문에서는 Vision Transformer(ViT) 기반 분류 알고리즘의 얼굴 연령 분류 성능을 비교 분석한다. ViT 기반분류 알고리즘으로는 최근 널리 사용되고 있는 ViT, Swin Transformer(ST), Neighborhood Attention Transformer(NAT) 세 가지로 선정하였으며, ViT의 얼굴 연령 분류 정확도 65.19%의 성능을 확인하였다.

• Journal of Korean Elementary Science Education
• /
• v.37 no.2
• /
• pp.126-146
• /
• 2018

The current study explores students' views about artificial intelligence (AI) through analyses of their drawings and perceptions. The data were gathered from a total of 177 elementary school students. The constant comparative analysis was used as the data analysis method. Based on the result, the current study found that students' views about artificial intelligence were constructed into two distinct dimensions: form and relationship. The form dimension, students' views about artificial intelligence were categorized into human, household goods, machine, smart computer, electronic chip/algorithm, or the hybridized form related to the game of go such as AlphaGo. On the relationship dimension, students' views about artificial intelligence were categorized into servants, friends or enemy. Given the combination of two dimensions, the current study found two noted patterns. The first, students who viewed artificial intelligence as human form perceived artificial intelligence as a friend or an enemy. However, those who viewed artificial intelligence as non-human form perceived artificial intelligence as a servant or an enemy. Based on these results, educational implications related to the preparation of artificial intelligence era for elementary science education are discussed.

• Journal of Korea Multimedia Society
• /
• v.23 no.12
• /
• pp.1486-1495
• /
• 2020

The analysis of digital microscopy images plays a vital role in computer-aided diagnosis (CAD) and prognosis. The main purpose of this paper is to develop a machine learning technique to predict the histological grades in prostate biopsy. To perform a multiclass classification, an AI-based deep learning algorithm, a multichannel convolutional neural network (MCCNN) was developed by connecting layers with artificial neurons inspired by the human brain system. The histological grades that were used for the analysis are benign, grade 3, grade 4, and grade 5. The proposed approach aims to classify multiple patterns of images extracted from the whole slide image (WSI) of a prostate biopsy based on the Gleason grading system. The Multichannel Convolution Neural Network (MCCNN) model takes three input channels (Red, Green, and Blue) to extract the computational features from each channel and concatenate them for multiclass classification. Stain normalization was carried out for each histological grade to standardize the intensity and contrast level in the image. The proposed model has been trained, validated, and tested with the histopathological images and has achieved an average accuracy of 96.4%, 94.6%, and 95.1%, respectively.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_1
• /
• pp.621-635
• /
• 2023

The assessment of structural condition is a crucial process for evaluating its usability and determining the diagnostic cycle. The currently employed manpower-based methods suffer from issues related to safety, efficiency, and objectivity. To address these concerns, research based on deep learning using images is being conducted. However, acquiring structural damage data is challenging, making it difficult to construct a substantial amount of training data, thus limiting the effectiveness of deep learning-based condition assessment. In this study, we propose a foundation model-based 2-step structural damage analysis to overcome the lack of training data in image-based structural condition assessments. We subdivided the elements of structural condition assessment into instantiation and quantification. In the quantification step, we applied a foundation model for image segmentation. Our method demonstrated a 10%-point increase in mean intersection over union compared to conventional image segmentation techniques, with a notable 40%-point improvement in the case of rebar exposure. We anticipate that our proposed approach will enhance performance in domains where acquiring training data is challenging.

• Journal of Korean Association for Spatial Structures
• /
• v.23 no.3
• /
• pp.95-103
• /
• 2023

With the increasing number of aging buildings across Korea, emerging maintenance technologies have surged. One such technology is the non-contact detection of concrete cracks via thermal images. This study aims to develop a technique that can accurately predict the depth of a crack by analyzing the temperature difference between the crack part and the normal part in the thermal image of the concrete. The research obtained temperature data through thermal imaging experiments and constructed a big data set including outdoor variables such as air temperature, illumination, and humidity that can influence temperature differences. Based on the collected data, the team designed an algorithm for learning and predicting the crack depth using machine learning. Initially, standardized crack specimens were used in experiments, and the big data was updated by specimens similar to actual cracks. Finally, a crack depth prediction technology was implemented using five regression analysis algorithms for approximately 24,000 data points. To confirm the practicality of the development technique, crack simulators with various shapes were added to the study.

• Smart Media Journal
• /
• v.11 no.7
• /
• pp.61-69
• /
• 2022

In the learned machine learning, the performance of machine learning degrades at the same time as drift occurs in terms of learning models and learning data over time. As a solution to this problem, I would like to propose the concept and evaluation method of ML drift to determine the re-learning period of machine learning. An XAI test and an XAI test of an apple image were performed according to strawberry and clarity. In the case of strawberries, the change in the XAI analysis of ML models according to the clarity value was insignificant, and in the case of XAI of apple image, apples normally classified objects and heat map areas, but in the case of apple flowers and buds, the results were insignificant compared to strawberries and apples. This is expected to be caused by the lack of learning images of apple flowers and buds, and more apple flowers and buds will be studied and tested in the future.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.232-234
• /
• 2022

Big data is widely used in customer relationship management, relationship marketing, financial business improvement, credit information and risk management. Moreover, as non-face-to-face financial transactions have become more active recently due to the COVID-19 virus, the use of financial big data is more demanded in terms of relationships with customers. In terms of customer relationship, financial big data has arrived at a time that requires an emotional rather than a technical approach. In relational marketing, it was necessary to emphasize the emotional aspect rather than the cognitive, rational, and rational aspects. Existing traditional financial data was collected and utilized through text-type customer transaction data, corporate financial information, and questionnaires. In this study, the customer's emotional image data, that is, atypical data based on the customer's cultural and leisure activities, is acquired through SNS and the customer's activity image is analyzed with an artificial intelligence CNN algorithm. Activity analysis is again applied to the annotated AI, and the AI big data model is designed to analyze the behavior model shown in the annotation.

• Korean Journal of Remote Sensing
• /
• v.38 no.1
• /
• pp.103-110
• /
• 2022

Since satellite images generally include clouds in the atmosphere, it is essential to detect or mask clouds before satellite image processing. Clouds were detected using physical characteristics of clouds in previous research. Cloud detection methods using deep learning techniques such as CNN or the modified U-Net in image segmentation field have been studied recently. Since image segmentation is the process of assigning a label to every pixel in an image, precise pixel-based dataset is required for cloud detection. Obtaining accurate training datasets is more important than a network configuration in image segmentation for cloud detection. Existing deep learning techniques used different training datasets. And test datasets were extracted from intra-dataset which were acquired by same sensor and procedure as training dataset. Different datasets make it difficult to determine which network shows a better overall performance. To verify the effectiveness of the cloud detection network such as Cloud-Net, two types of networks were trained using the cloud dataset from KOMPSAT-3 images provided by the AIHUB site and the L8-Cloud dataset from Landsat8 images which was publicly opened by a Cloud-Net author. Test data from intra-dataset of KOMPSAT-3 cloud dataset were used for validating the network. The simulation results show that the network trained with KOMPSAT-3 cloud dataset shows good performance on the network trained with L8-Cloud dataset. Because Landsat8 and KOMPSAT-3 satellite images have different GSDs, making it difficult to achieve good results from cross-sensor validation. The network could be superior for intra-dataset, but it could be inferior for cross-sensor data. It is necessary to study techniques that show good results in cross-senor validation dataset in the future.

• Korean Journal of Radiology
• /
• v.25 no.6
• /
• pp.518-539
• /
• 2024

Coronary computed tomography angiography (CCTA) has emerged as a pivotal tool for diagnosing and risk-stratifying patients with suspected coronary artery disease (CAD). Recent advancements in image analysis and artificial intelligence (AI) techniques have enabled the comprehensive quantitative analysis of coronary atherosclerosis. Fully quantitative assessments of coronary stenosis and lumen attenuation have improved the accuracy of assessing stenosis severity and predicting hemodynamically significant lesions. In addition to stenosis evaluation, quantitative plaque analysis plays a crucial role in predicting and monitoring CAD progression. Studies have demonstrated that the quantitative assessment of plaque subtypes based on CT attenuation provides a nuanced understanding of plaque characteristics and their association with cardiovascular events. Quantitative analysis of serial CCTA scans offers a unique perspective on the impact of medical therapies on plaque modification. However, challenges such as time-intensive analyses and variability in software platforms still need to be addressed for broader clinical implementation. The paradigm of CCTA has shifted towards comprehensive quantitative plaque analysis facilitated by technological advancements. As these methods continue to evolve, their integration into routine clinical practice has the potential to enhance risk assessment and guide individualized patient management. This article reviews the evolving landscape of quantitative plaque analysis in CCTA and explores its applications and limitations.

• The Journal of Bigdata
• /
• v.6 no.1
• /
• pp.37-49
• /
• 2021

The number of dementia patients in Korea is estimated to be over 800,000, and the severity of dementia is becoming a social problem. However, no treatment or drug has yet been developed to cure dementia worldwide. The number of dementia patients is expected to increase further due to the rapid aging of the population. Currently, early detection of dementia and delaying the course of dementia symptoms is the best alternative. This study presented a methodology for early diagnosis of dementia by measuring and analyzing amyloid plaques. This vital protein can most clearly and early diagnose dementia in the retina through AI-based image analysis. We performed binary classification and multi-classification learning based on CNN on retina data. We also developed a deep learning algorithm that can diagnose dementia early based on pre-processed retinal data. Accuracy and recall of the deep learning model were verified, and as a result of the verification, and derived results that satisfy both recall and accuracy. In the future, we plan to continue the study based on clinical data of actual dementia patients, and the results of this study are expected to solve the dementia problem.