• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.1
• /
• pp.1-7
• /
• 2023

Convolutional neural networks have already been applied to various fields beyond human visual processing capabilities in the image processing area. However, they are exposed to a severe risk of deteriorating model performance due to the appearance of adversarial attacks. In addition, defense technology to respond to adversarial attacks is effective against the attack but is vulnerable to other types of attacks. Therefore, to respond to an adversarial attack, it is necessary to analyze how the performance of the adversarial attack deteriorates through the process inside the convolutional neural network. In this study, the adversarial attack of the Alexnet and VGG11 models was analyzed using the population sparseness index, a measure of neuronal activity in neurophysiology. Through the research, it was observed in each layer that the population sparsity index for adversarial examples showed differences from that of benign examples.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.365-381
• /
• 2023

The existing vision-based techniques for inspection and condition assessment of civil infrastructure are mostly manual and consequently time-consuming, expensive, subjective, and risky. As a viable alternative, researchers in the past resorted to deep learning-based autonomous damage detection algorithms for expedited post-disaster reconnaissance of structures. Although a number of automatic damage detection algorithms have been proposed, the scarcity of labeled training data remains a major concern. To address this issue, this study proposed a semi-supervised learning (SSL) framework based on consistency regularization and cross-supervision. Image data from post-earthquake reconnaissance, that contains cracks, spalling, and exposed rebars are used to evaluate the proposed solution. Experiments are carried out under different data partition protocols, and it is shown that the proposed SSL method can make use of unlabeled images to enhance the segmentation performance when limited amount of ground truth labels are provided. This study also proposes DeepLab-AASPP and modified versions of U-Net++ based on channel-wise attention mechanism to better segment the components and damage areas from images of reinforced concrete buildings. The channel-wise attention mechanism can effectively improve the performance of the network by dynamically scaling the feature maps so that the networks can focus on more informative feature maps in the concatenation layer. The proposed DeepLab-AASPP achieves the best performance on component segmentation and damage state segmentation tasks with mIoU scores of 0.9850 and 0.7032, respectively. For crack, spalling, and rebar segmentation tasks, modified U-Net++ obtains the best performance with Igou scores (excluding the background pixels) of 0.5449, 0.9375, and 0.5018, respectively. The proposed architectures win the second place in IC-SHM2021 competition in all five tasks of Project 2.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.324-324
• /
• 2023

수제는 비용 대비 높은 효율을 가지는 수공 구조물로 하천과 해안에 많이 이용된다. 수제는 하천에서 식생물의 서식처, 유량 확보, 하천시설 보호 등의 목적으로 설치되며 해안에서는 해안선보호, 수심 확보 등의 다양한 목적으로 설치된다. 수제는 일반적으로 넓은 영역을 제어할 수 있도록 여러개의 군수제 형태로 설치한다. 수제설치 시 기존 하도나 수로의 흐름에 교란이 일어나는데 본류부에서는 하폭이 좁아져 수제선단부에 의한 영향으로 높은 유속이 발생되며 수제역내에서는 재순환흐름과 함께 다양한 형태의 유속분포가 형성된다. 이러한 수제에 의한 흐름은 수제 선단부의 영향으로 발생하는 흐름분리선(Separation Layer)을 기준으로 흐름중심선(Talweg Line)과 흐름분리영역(Separation area)을 만들어내며 수제역내의 재순환영역에서 완화된 유속은 제방을 침식으로부터 보호하며 식생물의 서식처를 제공한다. 기존의 연구는 대부분 단일수제에 대한 연구가 수행되었으며 연속된 군수제에 대한 연구는 다소 제한된 조건에서 수행되었다. 이에 본 연구에서는 연속된 수제가 설치된 경우 수리실험을 통해 수제주변의 흐름 변화 특성을 분석하였다. 수리실험은 30.0m(L')×1.52m(B')×1.10m(H')의 직선 개수로에 050m(l)×0.10m(b)×0.25m(h)의 수제를 수로측면을 따라 연속적으로 배열하여 설치하였다. 실험은 수심, 유량, 수제설치간격을 변화시키면서 수행했으며 마그네틱유속계를 사용해서 수제의 영향을 받지 않으며 벽면의 영향이 가장 적은 수로상류의 중앙지점에서 접근유속을 측정했다. 표면유속을 취득하기 위해 일반디지털카메라로 FHD(1920×1080) 해상도를 가진 영상을 각각의 조건마다 10~20분 촬영했다. 촬영된 영상을 연속된 정지영상으로 변환하고 영상해석 프로그램을 이용하여 표면유속을 도출했다. 표면영상유속계를 활용하여 도출한 결과로 각각 조건에 따라 발생하는 최대유속과 최대역방향유속, 와도(vorticity), 흐름중심선과 단면유속분포의 변화와 발생 위치를 분석했다.

• Restorative Dentistry and Endodontics
• /
• v.48 no.1
• /
• pp.1.1-1.11
• /
• 2023

Objectives: This study evaluated the dentinal penetration depth of 2.5% sodium hypochlorite (NaOCl) in root canals with and without preparation and different irrigant activation protocols. Materials and Methods: Sixty-three bovine mandibular incisors were randomly allocated to 6 groups (n = 10): G1, preparation + conventional needle irrigation (CNI); G2, preparation + passive ultrasonic irrigation (PUI); G3, preparation + Odous Clean (OC); G4, no preparation + CNI; G5, no preparation + PUI; G6, no preparation + OC; and CG (negative control; n = 3). Samples were filled with crystal violet for 72 hours. Irrigant activation was performed. Samples were sectioned perpendicularly along the long axis, 3 mm and 7 mm from the apex. Images of the root thirds of each block were captured with a stereomicroscope and analyzed with an image analysis software. One-way analysis of variance, followed by the Tukey post hoc test, and the Student's t-test were used for data analysis, with a significance level of 5%. Results: The NaOCl penetration depth was similar when preparation was performed, regardless of the method of irrigation activation (p > 0.05). In the groups without preparation, G6 showed greater NaOCl penetration depth (p < 0.05). The groups without preparation had a greater NaOCl penetration depth than those with preparation (p = 0.0019). Conclusions: The NaOCl penetration depth was similar in groups with root canal preparation. Without root canal preparation, OC allowed deeper NaOCl penetration. The groups without preparation had greater NaOCl penetration than those undergoing root canal preparation.

• Journal of Korea Entertainment Industry Association
• /
• v.13 no.8
• /
• pp.209-221
• /
• 2019

A found-footage film La France est notre patrie is a documentary, in which archive images are juxtaposed with intertitles, non-diegetic music and foley, by borrowing an audiovisual strategy of silent films. The filmmaker Rithy Panh has excavated the images, which had been taken during the same period as the film history of the end of the 19^th and early 20^th centuries in Southeast Asia and Africa under French colonial rule. This paper examines the filmmaker's methods of historiography when utilizing archive images in order to represent the past by referring to Walter Benjamin's concept of historical montage and dialectical image. As the analysis illustrates the singularity of constructive methods, which include multi-layer viewpoints and montage styles of compilation and collage, it reveals how La France est notre patrie elicits the essay film modes through its self-reflexivity, leads audience to the threshold of critical thinking about time and history and creates a discourse of counter-memory.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.249-251
• /
• 2022

In this paper, we present machine learning to control drone flight using EEG signals. We defined takeoff, forward, backward, left movement and right movement as control targets and measured EEG signals from the frontal lobe for controlling using Fp1. Fp2 Fp2 two-channel dry electrode (NeuroNicle FX2) measuring at 250Hz sampling rate. And the collected data were filtered at 6~20Hz cutoff frequency. We measured the motion image of the action associated with each control target open for 5.19 seconds. Using Matlab's classification learner for the measured EEG signal, the triple layer neural network, logistic regression kernel, nonlinear polynomial Support Vector Machine(SVM) learning was performed, logistic regression kernel was confirmed as the highest accuracy for takeoff and forward, backward, left movement and right movement of the drone in learning by class True Positive Rate(TPR).

• Journal of the Microelectronics and Packaging Society
• /
• v.30 no.1
• /
• pp.1-16
• /
• 2023

Recently, the shift to next-generation wide-bandgap (WBG) power semiconductor for electric vehicle is accelerated due to the need to improve power conversion efficiency and to overcome the limitation of conventional Si power semiconductor. With the adoption of WBG semiconductor, it is also required that the packaging materials for power modules have high temperature durability. As an alternative to conventional high-temperature Pb-based solder, Ag sintering die attach, which is one of the power module packaging process, is receiving attention. In this study, we will introduce the recent research trends on the Ag sintering die attach process. The effects of sintering parameters on the bonding properties and methodology on the exact physical properties of Ag sintered layer by the realization 3D image are discussed. In addition, trends in thermal shock and power cycle reliability test results for power module are discussed.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.6
• /
• pp.1-10
• /
• 2009

Flash memory has the advantages of nonvolatile, low power consumption, light weight, and high endurance. This enables the flash memory to be utilized as a storage of mobile computing device such as PMP(Portable Multimedia Player). Potable device with a mass flash memory can store various multimedia data such as video, audio, or image. Typical index systems for mobile computer are inefficient to search a form of text like lyric or title. In this paper, we propose a new text index system, named EMTEX(Embedded Text Index). EMTEX has the following salient features. First, it uses a compression algorithm for embedded system. Second, if a new insert or delete operation is executed on the base table. EMTEX updates the text index immediately. Third, EMTEX considers the characteristics of flash memory to design insert, delete, and rebuild operations on the text index. Finally, EMTEX is executed as an upper layer of DBMS. Therefore, it is independent of the underlying DBMS. We evaluate the performance of EMTEX. The Experiment results show that EMTEX can outperform th conventional index systems such as Oracle Text and FT3.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.36 no.2
• /
• pp.143-150
• /
• 2023

Underwater wireless communication is a challenging issue for realizing the smart aqua-farm and various marine activities for exploring the ocean and environmental monitoring. In comparison to acoustic and radio frequency technologies, the visible light communication is the most promising method to transmit data with a higher speed in complex underwater environments. To send data at a speedier rate, high-performance photodetectors are essentially required to receive blue and/or cyan-blue light that are transmitted from the light sources in a light-fidelity (Li-Fi) system. Here, we fabricated high-performance organic phototransistors (OPTs) based on P-type donor polymer (PTO2) and N-type acceptor small molecule (IT-4F) blend semiconductors. Bulk-heterojunction (BHJ) PTO2:IT-4F photo-active layer has a broad absorption spectrum in the range of 450~550 nm wavelength. Solution-processed OPTs showed a high photo-responsivity >1,000 mA/W, a large photo-sensitivity >10³, a fast response time, and reproducible light-On/Off switching characteristics even under a weak incident light. BHJ organic semiconductors absorbed photons and generated excitons, and efficiently dissociated to electron and hole carriers at the donor-acceptor interface. Printed and flexible OPTs can be widely used as Li-Fi receivers and image sensors for underwater communication and underwater internet of things (UIoTs).

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.6
• /
• pp.1516-1529
• /
• 2023

In this paper, a motor gearbox fault-detection system based on a hierarchical flow-based model is proposed. The proposed system is used for the anomaly detection of a motion sound-based actuator module. The proposed flow-based model, which is a generative model, learns by directly modeling a data distribution function. As the objective function is the maximum likelihood value of the input data, the training is stable and simple to use for anomaly detection. The operation sound of a car's side-view mirror motor is converted into a Mel-spectrogram image, consisting of a folding signal and an unfolding signal, and used as training data in this experiment. The proposed system is composed of an encoder and a decoder. The data extracted from the layer of the pretrained feature extractor are used as the decoder input data in the encoder. This information is used in the decoder by performing an interlayer cross-scale convolution operation. The experimental results indicate that the context information of various dimensions extracted from the interlayer hierarchical data improves the defect detection accuracy. This paper is notable because it uses acoustic data and a normalizing flow model to detect outliers based on the features of experimental data.