• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.649-651
• /
• 2018

본 논문에서는 CNN에서 클래스 활성화 맵과 원샷 러닝을 결합하여 트위터 분류를 위한 딥 러닝 모델을 제안한다. 클래스 활성화 맵은 트윗 분류에 대한 분류 주제와 연관된 핵심 어휘를 추출하고 강조 표시하도록 사용되었다. 특히 작은 학습 데이터 셋을 사용하여 다중 클래스 분류의 성능을 향상시키기 위해 원샷 러닝 방법을 적용한다. 제안하는 방법을 검증하기위해 TREC 2018 태스크의 사건 스트림(TREC-IS) 학습데이터를 사용하여 비교실험을 했다. 실험 결과에서 CNN 기본 모델의 정확도는 58.1%이고 제안 방법의 정확도는 69.6%로 성능이 향상됨을 보였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.307-310
• /
• 2021

X-ray 영상 검사·검출 문제에 기존 딥러닝 모델을 사용하려는 시도들이 존재해왔고, 합성곱 신경망의 강력한 표현력 덕분에 대체로 준수한 성능이 보장되었다. 그러나 문제의 특성에 따라 기대한 만큼의 분류 및 검출 성능이 나오지 않는 경우가 존재한다. 이는 1) 검출 대상의 스케일이 다양하거나, 2) X-ray 영상은 흑백 영상으로 미세한 특징을 학습하기 어렵거나, 3) 지도학습을 하기에는 학습 데이터의 양이 부족하기 때문인 것이 주요 원인들이다. 본 논문에서는 다양한 스케일의 특징맵을 추출하여 종합적으로 학습하는 신경망을 통해, '생선살 X-ray 영상' 데이터셋에서 '생선 가시' 이물질 class가 모델 내에서 어떻게 학습되는지를 살펴본다. 그리고 X-ray 영상의 경우, 이물질 class를 크기별로 새롭게 labeling하여 성능 개선이 일어날 수 있음을 보인다. 또한 Multi-scale CAM을 통해 class에 따른 활성화 정도를 시각화하여 모델을 직관적으로 분석할 수 있음을 보일 것이다.

• Journal of Korea Spatial Information System Society
• /
• v.11 no.1
• /
• pp.157-168
• /
• 2009

Geospatial information, which deeply has an effect on our life, have been evolved as intelligent geospatial information in Ubiquitous era. Also, Various services are introduced using the intelligent geospatial information. However, there is no classification system, for understanding the intelligent geospatial information services, considering any developers and users. It needs to be classification system to classify these services. In this paper, we introduced a concept of intelligent geospatial information and developed a service classification matrix regarding to the features of the services. This service classification matrix has three scales; service domain, service intelligent level, and geo-location accuracy. The propose of this matrix can be utilized in two aspects. First, the matrix can improve the reality that doesn't reflect actual demands for the services. Second, the matrix can present the goal of the new services or the development direction. The matrix can be utilized to the geospatial industry as creating the new blue ocean services. However, the service classification matrix needs to modify and complement to have no anything wrong when the various services are applied to the matrix. In the long run, the matrix has to be utilized as a material to make out a service roadmap or TRM(Technical Reference Model).

• Journal of Energy Engineering
• /
• v.27 no.3
• /
• pp.21-27
• /
• 2018

In accordance with the implementation of the Zero Energy Building Certification System, it for the activation and expansion of the private sector is being steadily upgraded. Also The government has set up a step-by-step mandatory roadmap until it is expanded to the private sector, starting with the public sector. We analyzed the energy requirements of existing buildings from 2016 to 2017 and the by load relationships of major factor. Of the existing buildings, 714 buildings in central and southern areas excluding residential buildings such as apartments and officetels were classified and their primary energy requirements were analyzed. As new design technologies are applied, the demand for energy from the passive side is steadily declining. In addition, there is a need to interpret various methods to improve the zero energy building certification standard in the point that the zero energy building pilot project is continuously carried out in relation to the activation of renewable energy supply.

• Journal of radiological science and technology
• /
• v.45 no.1
• /
• pp.41-47
• /
• 2022

The purpose of this study is to apply a deep learning model that can distinguish lung perfusion and lung ventilation images in nuclear medicine, and to evaluate the image classification ability. Image data pre-processing was performed in the following order: image matrix size adjustment, min-max normalization, image center position adjustment, train/validation/test data set classification, and data augmentation. The convolutional neural network(CNN) structures of VGG-16, ResNet-18, Inception-ResNet-v2, and SE-ResNeXt-101 were used. For classification model evaluation, performance evaluation index of classification model, class activation map(CAM), and statistical image evaluation method were applied. As for the performance evaluation index of the classification model, SE-ResNeXt-101 and Inception-ResNet-v2 showed the highest performance with the same results. As a result of CAM, cardiac and right lung regions were highly activated in lung perfusion, and upper lung and neck regions were highly activated in lung ventilation. Statistical image evaluation showed a meaningful difference between SE-ResNeXt-101 and Inception-ResNet-v2. As a result of the study, the applicability of the CNN model for lung scintigraphy classification was confirmed. In the future, it is expected that it will be used as basic data for research on new artificial intelligence models and will help stable image management in clinical practice.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.27 no.2
• /
• pp.363-368
• /
• 2021

Ship engine rooms have improved automation systems owing to the advancement of technology. However, there are many variables at sea, such as wind, waves, vibration, and equipment aging, which cause loosening, cutting, and leakage, which are not measured by automated systems. There are cases in which only one engineer is available for patrolling. This entails many risk factors in the engine room, where rotating equipment is operating at high temperature and high pressure. When the engineer patrols, he uses his five senses, with particular high dependence on vision. We hereby present a preliminary study to implement an engine-room patrol robot that detects and informs the machine room while a robot patrols the engine room. Images of ship engine-room equipment were classified using a convolutional neural network (CNN). After constructing the image dataset of the ship engine room, the network was trained with a pre-trained CNN model. Classification performance of the trained model showed high reproducibility. Images were visualized with a class activation map. Although it cannot be generalized because the amount of data was limited, it is thought that if the data of each ship were learned through transfer learning, a model suitable for the characteristics of each ship could be constructed with little time and cost expenditure.

• Journal of KIISE:Computing Practices and Letters
• /
• v.11 no.3
• /
• pp.235-245
• /
• 2005

The rapid development of computer network technology has brought the Internet as the major infrastructure to our society. But the rapid increase in malicious computer intrusions using such technology causes urgent problems of protecting our information society. The recent trends of the intrusions reflect that the intruders do not break into victim host directly and do some malicious behaviors. Rather, they tend to use some automated intrusion tools to penetrate systems. Most of the unknown types of the intrusions are caused by using such tools, with some minor modifications. These tools are mostly similar to the Previous ones, and the results of using such tools remain the same as in common patterns. In this paper, we are describing design and implementation of attacker-traceback system, which traces the intruder based on the multi-agent architecture. The system first applied SOM to classify the unknown types of the intrusion into previous similar intrusion classes. And during the intrusion analysis stage, we formalized the patterns of the tools as a knowledge base. Based on the patterns, the agent system gets activated, and the automatic tracing of the intrusion routes begins through the previous attacked host, by finding some intrusion evidences on the attacked system.

• Journal of the Korea Computer Graphics Society
• /
• v.29 no.3
• /
• pp.33-43
• /
• 2023

Not only emotions affect physical sensations, but they also have an impact on physical movements. The responses to emotions vary depending on the type of emotional stimuli. However, research on the effects of emotional stimuli on the activation of bodily movements has not been rigorously examined, and these effects have not been investigated in Autonomous Emotion Recognition (AER) systems. In this study, we aimed to compare the emotional responses of 20 participants to three types of emotional stimuli (words, pictures, and videos) and investigate their activation or deactivation for the AER system. Our dependent measures included emotional responses, computer-based self-reporting methods, and bodily movements recorded using motion capture devices. The results suggested that video stimuli elicited higher levels of emotional movement, and emotional movement patterns were similar across different types of emotional stimuli for happiness, sadness, anger, and neutrality. Additionally, the findings indicated that bodily changes observed during video stimuli had the highest classification accuracy. These findings have implications for future research on the bodily changes elicited by emotional stimuli.