• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1561-1567
• /
• 2015

Abstract Semi-supervised learning is an area in machine learning that employs both labeled and unlabeled data in order to train a model and has the potential to improve prediction performance compared to supervised learning. Graph-based semi-supervised learning has recently come into focus with two phases: graph construction, which converts the input data into a graph, and label inference, which predicts the appropriate labels for unlabeled data using the constructed graph. The inference is based on the smoothness assumption feature of semi-supervised learning. In this study, we propose an enhanced label inference algorithm by incorporating the importance of each vertex. In addition, we prove the convergence of the suggested algorithm and verify its excellence.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.659-662
• /
• 2021

국내에서는 매년 많은 수의 자격시험이 치러지고 있다. 현재 대부분의 시험장에서 응시자 본인 확인 절차는 감독관이 응시자의 얼굴과 신분증 사진을 비교하는 방식으로 이루어진다. 하지만 이 방식은 사람에 따라 오차가 클 수 있으며, 사진과 눈에 띄는 차이가 없으면 동일인물로 판단하기 쉽다. 최근까지도 대리응시 이슈가 발생하고 있어 근절을 위한 보다 강력한 조치가 필요하다. 본 논문에서는 지문과 오토인코더, SGAN을 이용하여 대리응시방지를 강화할 수 있는 본인 확인 시스템을 제안한다. 이때 응시자의 지문정보가 그대로 인증 서버에 저장되면 응시자의 생체정보가 노출될 수 있다는 문제점이 존재한다. 따라서 오토인코더를 통해 지문의 특징점만 추출하여 인증용 이미지를 생성하고 이 이미지를 서버에 저장하여 학습시키도록 한다. 적은 학습데이터 환경에서 분류기로써 좋은 성능을 갖는 SGAN을 통해 인증 이미지와 응시자가 동일인물인지 확인할 수 있다. 서버가 공격을 받더라도 응시자의 지문데이터가 그대로 노출되지 않게 되어 보안 취약점을 극복할 수 있다.

• Journal of information and communication convergence engineering
• /
• v.21 no.1
• /
• pp.98-102
• /
• 2023

This paper presents a study on how augmenting semi-synthetic image data improves the performance of human detection algorithms. In the field of object detection, securing a high-quality data set plays the most important role in training deep learning algorithms. Recently, the acquisition of real image data has become time consuming and expensive; therefore, research using synthesized data has been conducted. Synthetic data haves the advantage of being able to generate a vast amount of data and accurately label it. However, the utility of synthetic data in human detection has not yet been demonstrated. Therefore, we use You Only Look Once (YOLO), the object detection algorithm most commonly used, to experimentally analyze the effect of synthetic data augmentation on human detection performance. As a result of training YOLO using the Penn-Fudan dataset, it was shown that the YOLO network model trained on a dataset augmented with synthetic data provided high-performance results in terms of the Precision-Recall Curve and F1-Confidence Curve.

• Advanced Industrial SCIence
• /
• v.2 no.4
• /
• pp.28-35
• /
• 2023

In this paper, a diffractive optical element was designed to create lighting patterns that satisfy the requirements of adaptive headlight systems for normal road mode, highway mode, and wet road mode, and this was rendered into a GDSII stream format file.To verify the effectiveness of the light distribution formed by the diffractive optical elements and the realization of white light, simulations based on Field Tracing and Ray Tracing were conducted, confirming the satisfaction of position and luminance requirements at the transformation beam measurement points. Based on this research, it is anticipated that the implementation of adaptive headlights would be possible, enabling the reproduction of luminance contrast and the creation of a simple-structured adaptive headlight system.

• Korean Journal of Remote Sensing
• /
• v.29 no.6
• /
• pp.601-610
• /
• 2013

As a standard water clarity variable, the vertical underwater visibility, called Secchi depth, is estimated with ocean color satellite data. In the present study, Moderate Resolvtion Imaging Spectradiometer (MODIS) data are used to measure the Secchi depth which is a useful indicator of ocean transparency for estimating the water quality and productivity. To estimate the Secchi depth $Z_v$, the empirical regression model is developed based on the satellite optical data and in-situ data. In the previous study, a semi-analytical algorithm for estimating $Z_v$ was developed and validated for Case 1 and 2 waters in both coastal and oceanic waters using extensive sets of satellite and in-situ data. The algorithm uses the vertical diffuse attenuation coefficient, $K_d$($m^{-1}$) and the beam attenuation coefficient, c($m^{-1}$) obtained from satellite ocean color data to estimate $Z_v$. In this study, the semi-analytical algorithm is validated using temporal MODIS data and in-situ data over the Yellow, Southern and East Seas including Case 1 and 2 waters. Using total 156 matching data, MODIS $Z_v$ data showed about 3.6m RMSE value and 1.7m bias value. The $Z_v$ values of the East Sea and Southern Sea showed higher RMSE than the Yellow Sea. Although the semi-analytical algorithm used the fixed coupling constant (= 6.0) transformed from Inherent Optical Properties (IOP) and Apparent Optical Properties (AOP) to Secchi depth, various coupling constants are needed for different sea types and water depth for the optimum estimation of $Z_v$.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.5
• /
• pp.17-28
• /
• 2023

Potholes that occur on paved roads can have fatal consequences for vehicles traveling at high speeds and may even lead to fatalities. While manual detection of potholes using human labor is commonly used to prevent pothole-related accidents, it is economically and temporally inefficient due to the exposure of workers on the road and the difficulty in predicting potholes in certain categories. Therefore, completely preventing potholes is nearly impossible, and even preventing their formation is limited due to the influence of ground conditions closely related to road environments. Additionally, labeling work guided by experts is required for dataset construction. Thus, in this paper, we utilized the Mean Teacher technique, one of the semi-supervised learning-based knowledge distillation methods, to achieve robust performance in pothole image classification even with limited labeled data. We demonstrated this using performance metrics and GradCAM, showing that when using semi-supervised learning, 15 pre-trained CNN models achieved an average accuracy of 90.41%, with a minimum of 2% and a maximum of 9% performance difference compared to supervised learning.

• Kyungpook Mathematical Journal
• /
• v.51 no.2
• /
• pp.205-215
• /
• 2011

Given a knot diagram D, we construct a semi-threading circle of it which can be an axis of D as a closed braid depending on knot diagrams. In particular, we consider semi-threading circles of minimal diagrams of a knot with respect to overpasses which give us some information related to the braid index. By this notion, we try to give another proof of the fact that, for every nontrivial knot K, the braid index b(K) of K is not less than the minimum number l(K) of overpasses of diagrams. Also, they are the same for a torus knot.

• 한국정보컨버전스학회:학술대회논문집
• /
• 2008.06a
• /
• pp.93-97
• /
• 2008

In this letter, a novel wavelet-based semi-fragile watermarking scheme is presented which exploiting the time-frequency feature of chaotic map. We also analyze the robustness to mild modification and fragility to malicious attack of our scheme. Its application includes tamper detection, image verification and copyright protection of multimedia content. Simulation results show the scheme can detect and localize malicious attacks with high peak signal-to-noise ratio(PSNR), while tolerating certain degree of JPEG compression and channel additive white Gaussian noise(AWGN)

• East Asian mathematical journal
• /
• v.35 no.1
• /
• pp.59-66
• /
• 2019

In this paper we study implicit-explicit (IMEX) methods combined with a semi-Lagrangian scheme to evaluate the prices of fixed strike arithmetic Asian options under jump-diffusion models. An Asian option is described by a two-dimensional partial integro-differential equation (PIDE) that has no diffusion term in the arithmetic average direction. The IMEX methods with the semi-Lagrangian scheme to solve the PIDE are discretized along characteristic curves and performed without any fixed point iteration techniques at each time step. We implement numerical simulations for the prices of a European fixed strike arithmetic Asian put option under the Merton model to demonstrate the second-order convergence rate.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.3
• /
• pp.1403-1417
• /
• 2019

With the rapid growth of multimedia data, research on cross-media feature learning has significance in many applications, such as multimedia search and recommendation. Existing methods are sensitive to noise and edge information in multimedia data. In this paper, we propose a semi-supervised method for cross-media feature learning by means of $L_{2,q}$ norm to improve the performance of cross-media retrieval, which is more robust and efficient than the previous ones. In our method, noise and edge information have less effect on the results of cross-media retrieval and the dynamic patch information of multimedia data is employed to increase the accuracy of cross-media retrieval. Our method can reduce the interference of noise and edge information and achieve fast convergence. Extensive experiments on the XMedia dataset illustrate that our method has better performance than the state-of-the-art methods.