• Review of KIISC
• /
• v.29 no.6
• /
• pp.53-58
• /
• 2019

최근 5G 네트워크의 발전으로 사물인터넷의 활용도가 커지며 시장이 급격히 확대되고 있다. 사물인터넷 기기가 급증하면서 이를 대상으로 하는 위협이 크게 늘며 사물인터넷 기기의 보안이 중요시 되고 있다. 그러나 이러한 사물인터넷 기기는 기존의 ICT 장비와는 다르게 리소스가 제한되어 있다. 본 논문에서는 이러한 특성을 갖는 사물인터넷 환경에 적합한 보안기술로 네트워크 학습을 통해 사물인터넷 기기의 이상행위를 탐지하는 경량화된 인공신경망 기술을 제안한다. 기기 별 혹은 사용자 별 네트워크 행위 패턴을 분석하여 특성 연구를 진행하였으며, 사물인터넷 기기의 정상행위를 수집하고 학습데이터로 활용한다. 이러한 학습데이터를 통해 인공신경망 기반의 오토인코더 알고리즘을 활용하여 이상행위 탐지 모델을 구축하였으며, 파라미터 튜닝을 통해 모델 사이즈, 학습 시간, 복잡도 등을 경량화 하였다. 본 논문에서 제안하는 탐지 모델은 신경망 프루닝 및 양자화를 통해 경량화된 오토인코더 기반 인공신경망을 학습하였으며, 정상 행위 패턴을 벗어나는 이상행위를 식별할 수 있었다. 본 논문은 1. 서론을 통해 현재 사물인터넷 환경과 보안 기술 연구 동향을 소개하고 2. 관련 연구를 통하여 머신러닝 기술과 이상 탐지 기술에 대해 소개한다. 3. 제안기술에서는 본 논문에서 제안하는 인공신경망 알고리즘 기반의 사물인터넷 이상행위 탐지 기술에 대해 설명하고, 4. 향후연구계획을 통해 추후 활용 방안 및 고도화에 대한 내용을 작성하였다. 마지막으로 5. 결론을 통하여 제안기술의 평가와 소회에 대해 설명하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.918-919
• /
• 2023

딥러닝 기술이 발전함에 따라 모델의 복잡성 역시 증가하고 있다. 본 연구에서는 모델 경량화를 위해 Knowledge Distillation 기법을 Crowd Counting Model에 적용했다. M-SFANet을 Teacher 모델로, 파라미터수가 적은 MCNN 모델을 Student 모델로 채택해 Knowledge Distillation을 적용한 결과, 기존의 MCNN 모델보다 성능을 향상했다. 이는 정확도와 메모리 효율성 측면에서 많은 개선을 이루어 컴퓨팅 리소스가 부족한 기기에서도 본 모델을 실행할 수 있어 많은 활용이 가능할 것이다.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2019.05a
• /
• pp.248-249
• /
• 2019

In the nuclear industry, in order to keep pace with the 4th industrial revolution era, they are trying to improve the construction and maintenance ability by utilizing the technologies such as digital twin and VR/AR from the construction stage. However, the nuclear 3D CAD model, which is used as the base in the latest technology, is heavy due to a large number of facilities per unit space compared to other industrial companies, and it is difficult to directly incorporate the latest technology into the results of CAD programs for design purposes. In this study, in order to improve usability, we tried to lighten the 3D model. First, we analyze the existing nuclear power plant 3D model and draw out the problems and features. Secondly, we derived the factors to consider when we make the 3D CAD models lightweight.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.900-901
• /
• 2023

본 연구는 TFLite 기반의 경량화 AI 모델을 활용하여 쓰레기의 자동 분리수거 시스템을 구축하는 방법을 제안한다. 제안된 시스템은 객체 인식 기술을 활용해 쓰레기를 정확하게 분류하며, 테스트 결과 평균 90.33%의 mAP 성능을 나타낸다. Label 수와 데이터셋의 한계가 존재하지만, 본 연구를 확장하고 개선함으로써 자동 분리수거의 효율성을 더욱 높일 수 있을 것으로 기대된다.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.8
• /
• pp.379-384
• /
• 2017

In real software business environment, there are required to validate software quality in diverse usage range of software for many small & medium companies. Software quality means both qualities of production and process. In our situation, we focus on better process quality of a test organization than a whole organization. But even the original test maturity model (TMM) does not enough to apply with our domestic venture/small & medium companies. To solve this problem, we suggest the simplified test maturity model for our companies. We redefine this simplified model with the original TMM and a test process improvement next (TPI next) model. The previous models just have provided each definition of maturity level, goal and activity per each level, which not exists an assessment guideline and a formal assessing procedure. Due to this reasons, an assessor is difficult to assess the test organization without them. this paper suggest an assessment guideline of the simplified TMM and also define the procedure which is included with activities and byproducts. With these assessment guideline, an assessor can work possible to formally assess test organizations of small & medium companies, and with self assessment guideline they can be correctly provision before assessment of their organizations.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.80-81
• /
• 2021

딥 러닝 모델 사용에 있어서, 일반적인 사용자가 이용할 수 있는 하드웨어 리소스는 제한적이기 때문에 기존 모델을 경량화 할 수 있는 프루닝 방법을 통해 제한적인 리소스를 효과적으로 활용할 수 있도록 한다. 그 방법으로, 여러 딥 러닝 모델들 중 비교적 파라미터 수가 많은 것으로 알려진 GAN 아키텍처에 네트워크 프루닝을 적용함으로써 비교적 무거운 모델을 적은 파라미터를 통해 학습할 수 있는 방법을 제시한다. 또한, 본 논문을 통해 기존의 SRGAN 논문에서 가장 효과적인 결과로 제시했던 16 개의 residual block 의 개수를 실제로 줄여 봄으로써 기존 논문에서 제시했던 결과와의 차이에 대해 서술한다.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.25 no.1
• /
• pp.65-71
• /
• 2012

In order to improve the structural safety and light-weight design of aircraft floats, natural frequency and static stress analysis are performed under water and ground landing conditions. A finite element mesh based on the design configuration of light aircraft floats is modeled, and simplified water and ground landing loads are applied to this model. The natural frequency and stress analysis of aluminum-alloy floats are carried out first. Then, the structural performance of the floats is re-analyzed in the case of laminated composites, and the numerical results are compared each other. It is concluded that, by tailoring the laminated composites with respect to stacking sequence and ply thickness, the structural safety of the light-weight floats can be improved.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.10
• /
• pp.17-25
• /
• 2023

In this paper, we propose a lightweight deep learning architecture tailored for efficient text detection in fashion design sketch images. Given the increasing prominence of Digital Transformation in the fashion industry, there is a growing emphasis on harnessing digital tools for creating fashion design sketches. As digitization becomes more pervasive in the fashion design process, the initial stages of text detection and recognition take on pivotal roles. In this study, a lightweight network was designed by building upon existing text detection deep learning models, taking into consideration the unique characteristics of apparel design drawings. Additionally, a separately collected dataset of apparel design drawings was added to train the deep learning model. Experimental results underscore the superior performance of our proposed deep learning model, outperforming existing text detection models by approximately 20% when applied to fashion design sketch images. As a result, this paper is expected to contribute to the Digital Transformation in the field of clothing design by means of research on optimizing deep learning models and detecting specialized text information.

• Smart Media Journal
• /
• v.12 no.5
• /
• pp.28-35
• /
• 2023

In this study, we propose a new light-weight model RoutingConvNet with fewer parameters to improve the applicability and practicality of speech emotion recognition. To reduce the number of learnable parameters, the proposed model connects bidirectional MFCCs on a channel-by-channel basis to learn long-term emotion dependence and extract contextual features. A light-weight deep CNN is constructed for low-level feature extraction, and self-attention is used to obtain information about channel and spatial signals in speech signals. In addition, we apply dynamic routing to improve the accuracy and construct a model that is robust to feature variations. The proposed model shows parameter reduction and accuracy improvement in the overall experiments of speech emotion datasets (EMO-DB, RAVDESS, and IEMOCAP), achieving 87.86%, 83.44%, and 66.06% accuracy respectively with about 156,000 parameters. In this study, we proposed a metric to calculate the trade-off between the number of parameters and accuracy for performance evaluation against light-weight.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.37 no.9
• /
• pp.1099-1107
• /
• 2013

This study proposes a weight reduction design for urban transit, specifically, a Korean EMU carbody made of aluminum extrusion profiles, according to size optimization and useful material changes. First, the thickness of the under-frame, side-panels, and end-panels were optimized by the size optimization process, and then, the weight of the Korean EMU carbody could be reduced to approximately 14.8%. Second, the under-frame of the optimized carbody was substituted with a frame-type structure made of SMA 570, and then, the weight of the hybrid-type carbody was 3.8% lighter than that of the initial K-EMU. Finally, the under-frame and the roof-panel were substituted with a composite material sandwich to obtain an ultralight hybrid-type carbody. The weight of the ultralight hybrid-type carbody was 30% lighter than that of the initial K-EMU. All the resulting carbody models satisfied the design regulations of the domestic Performance Test Standard for Electrical Multiple Unit.