• The Journal of Korea Robotics Society
• /
• v.12 no.3
• /
• pp.313-321
• /
• 2017

One of the most frequently performed tasks in human-robot interaction (HRI), intelligent vehicles, and security systems is face related applications such as face recognition, facial expression recognition, driver state monitoring, and gaze estimation. In these applications, accurate head pose estimation is an important issue. However, conventional methods have been lacking in accuracy, robustness or processing speed in practical use. In this paper, we propose a novel method for estimating head pose with a monocular camera. The proposed algorithm is based on a deep neural network for multi-task learning using a small grayscale image. This network jointly detects multi-view faces and estimates head pose in hard environmental conditions such as illumination change and large pose change. The proposed framework quantitatively and qualitatively outperforms the state-of-the-art method with an average head pose mean error of less than $4.5^{\circ}$ in real-time.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.12
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• Proceedings of the KSME Conference
• /
• 2001.06e
• /
• pp.57-62
• /
• 2001

Korea Atomic Energy Research Institute (KAERI) launched an intermediate scale steam explosion experiment named Test for Real cOrium Interaction with water (TROI) using reactor material to investigate whether the corium would lead to energetic steam explosion when interacted with cold water at low pressure. The melt-water interaction is confined in a pressure vessel with the multi-dimensional fuel and water pool geometry. The cold crucible technology, where the mixture of powder in a water-cooled cage is heated by high frequency induction, is employed. In this paper, results of the first series of tests ($TROI-1{\sim}5$) were discussed. The ZrO2 jets with 5kg mass and 5cm diameter were poured into the 67cm deep water pool at $30{\sim}95^{\circ}C$. Either spontaneous steam explosions or quenching was observed. The morphology of debris and pressure wave profiles clearly indicates the each case.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.37 no.5
• /
• pp.433-441
• /
• 2009

Generally, the events in nature have multi-disciplinary characteristics. To solve this problems, these days loosely coupled methods are widely applied because of advantage of solvers which are already developed and well proved. Those solvers use different mesh system, so transformation and mapping of data are vital in the field of fluid-structure interaction(FSI). In this paper, the interpolation of deformation which is used globally and compactly supported radial basis functions(RBF), and mapping of force which use principle of virtual work are examined for computing time and accuracy to compare ability with simple 3-D problem. As the results, interpolation scheme of compactly supported radial basis functions are useful to interpolation and mapping for large-scale airplane in FSI with a k-dimensional tree(kd-tree) which is a space-partitioning data structure for organizing points in a k-dimensional space.

• Korean Journal of Child Education & Care
• /
• v.17 no.2
• /
• pp.99-114
• /
• 2017

This study investigated the affects of Parent-child interaction, peer interaction and teacher-child interaction on kindergartners' social competence, language and literacy Ability. Participants were 1203 children attending panel study on Korean children. It used the 'Social Competence Inventory', 'SECCYD', 'HEQ', 'PIPPS' and the 'Teacher-Child Relation Scale' in collection of data. Collected data were analyzed through the SPSS 18.0 program and pearson's correlation and step-wise multi regression analysis. Findings are as follows. First, there were positive correlations between children's social competence and parent-child interaction, peer interaction and teacher-child interaction. Second, there were positive correlations between children's language and literacy ability and peer interaction and teacher-child interaction. Third, Parent-child interaction, peer interaction were predictable variable affecting the young children's social competence. Fourth, peer interaction and teacher-child interaction were found to be predictable variables affecting the young children's language and literacy ability.

• Journal of Distribution Research
• /
• v.10 no.1
• /
• pp.33-57
• /
• 2005

After 1990's, many multi-level marketing companies lave been introduced in Korea. These MLM companies operate their stores as same way as general retailing stores. The major characteristics of these MLM companys' store is that their main customers are distributors who sell the purchased merchandise to another customers. Many studies about merchandise display in general retailing stores were reported. But, there was less research about merchandise display in these special type stores such as MLM companys' stores. This paper investigates the effects of merchandise display on distributor's merchandise selection in channel flow of multi-level marketing company. For this purpose, we formulated four hypotheses about display variance in quantity, height, location, and related merchandise to analyze the effect of merchandise display methods in MLM companys' stores. The experiment had been proceeded in three stores in a MLM company for 6 weeks and the sales data were collected by POS. The methods to analyze the data were used ANOVA and T-test. Findings of this study are as follows; First, there was no effect to store sale by the interaction effect between merchandise display method and scale of store. Second, scale of store affected considerably the volume of sales of each store according to main effect analysis. Third, display variance in quantity, height, and location did not affect store sale. In the related merchandise display, however, sale in all store was increased. Fourth, in additional analysis considering merchandise display only, display variance in both height and location affected their sale in large scale store. Based on the above results, we may predict merchandise display can affect sale in MLM companys' store as well as general retailing stores. Therefore if MLM company has large scale store, it should consider merchandise display methods in its stores.

• Journal of the Korea Computer Graphics Society
• /
• v.29 no.3
• /
• pp.127-135
• /
• 2023

Accurate segmentation of histopathology whole slide images (WSIs) is a crucial task for disease diagnosis and treatment planning. However, conventional automated segmentation algorithms may not always be applicable to WSI segmentation due to their large size and variations in tissue appearance, staining, and imaging conditions. Recent advances in interactive segmentation, which combines human expertise with algorithms, have shown promise to improve efficiency and accuracy in WSI segmentation but also presented us with challenging issues. In this paper, we propose a novel interactive segmentation method, ZoomISEG, that leverages multi-resolution WSIs. We demonstrate the efficacy and performance of the proposed method via comparison with conventional single-scale methods and an ablation study. The results confirm that the proposed method can reduce human interaction while achieving accuracy comparable to that of the brute-force approach using the highest-resolution data.

• The Journal of the Korea Contents Association
• /
• v.20 no.3
• /
• pp.452-461
• /
• 2020

The purpose of this study is to identify how learners perceive mobile technology-based learning materials. For this purpose, two methods were utilized. Using multi-dimensional scale(MDS), it was identified that how learners perceive each type of learning materials using mobile technology. Through semantic differential scale(SDS), learners' perception of the difference between mobile learning materials and existing traditional learning materials was analyzed. As a result, learning materials using mobile technology were classified into as follows : the dimension of interaction with the content; the sense of presence. Learners perceived that mobile learning materials had characteristics of 'active', 'learner-centric', 'multi-sensory', and 'stimulating interest'. The significance of this study was to empirically and comprehensively investigate learners' perception for the characteristics of various mobile learning materials.

• Proceedings of the Korean Vacuum Society Conference
• /
• 1998.02a
• /
• pp.135-136
• /
• 1998

By the ion bombardment the original discrete layered structure is damaged and a uniformly mixed layer is formed by the intermixing of the films. Immediately after this dynamic cascade mixing a structure of this mixed layer is likely to be a mixture of randomly distributed atoms. Subsequently the mixed layered structure becomes a non-equilibrium structure such as the metastable pphase because the kinetic energies of the incident ions rappidly dissippate and host atoms within the collision cascade region are quenched from a highly energetic state. The formation of the metastable transition metal alloys using ion-beam-mixing has been extensively studied for many years because of their sppecific ppropperties that differ from those of bulk materials. in ion-beam-mixing the alloy or comppound is formed due to the atomic interaction between different sppecies during ion bombardment. in this study the metastable pphase formed by ion-beam-mixing pprocess is comppared with equilibrium one by arc-melting method by GXRD and XAS. Therfore we studied the fundamental characteristics of charge redistribution uppon alloying and formation of intermetallic comppounds. The multi-layer films were depposited on a wet-oxidized Si(100) substrate by sequential electron beam evapporation at a ppressure of less than 5$\times$10-7 Torr during depposition. These compprise 4 ppairs of Co and ppt layers where thicknesses of each layer were varied in order to change the alloy compposition.

• Ocean Systems Engineering
• /
• v.7 no.3
• /
• pp.299-318
• /
• 2017

The paper aims at illustrating several key issues and ongoing efforts for development of a reliable fully-Lagrangian particle-based solver for simulation of hydroelastic slamming. Fluid model is founded on the solution of Navier-Stokes along with continuity equations via an enhanced version of a projection-based particle method, namely, Moving Particle Semi-implicit (MPS) method. The fluid model is carefully coupled with a structure model on the basis of conservation of linear and angular momenta for an elastic solid. The developed coupled FSI (Fluid-Structure Interaction) solver is applied to simulations of high velocity impact of an elastic aluminum wedge and hydroelastic slammings of marine panels. Validations are made both qualitatively and quantitatively in terms of reproduced pressure as well as structure deformation. Several remaining challenges as well as important key issues are highlighted. At last, a recently developed multi-scale MPS method is incorporated in the developed FSI solver towards enhancement of its adaptivity.