• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.386-388
• /
• 2022

Recently, realistic contents using virtual reality and augmented reality are attracting attention as learning aids. 3D-based contents have the advantage of being able to observe and experience objects from various angles than 2D-based contents shown on a flat surface. In this paper, we propose a content design based on 3D model for Arduino learning in a virtual environment. The Arduino board and sensor were implemented using Blender, and a 3D-based simulator environment was constructed using the Unity engine. The proposed content uses the Arduino board and sensor implemented in 3D so that learners can easily experience the working principle of Arduino and the coding process.

• Journal of the Korea Computer Graphics Society
• /
• v.26 no.4
• /
• pp.1-8
• /
• 2020

Motion synthesis using physics-based controllers can generate a character animation that interacts naturally with the given environment and other characters. Recently, various methods using deep neural networks have improved the quality of motions generated by physics-based controllers. In this paper, we present a control policy learned by deep reinforcement learning (DRL) that enables Luxo, the mascot character of Pixar animation studio, to run towards a random goal location while imitating a reference motion and maintaining its balance. Instead of directly training our DRL network to make Luxo reach a goal location, we use a reference motion that is generated to keep Luxo animation's jumping style. The reference motion is generated by linearly interpolating predetermined poses, which are defined with Luxo character's each joint angle. By applying our method, we could confirm a better Luxo policy compared to the one without any reference motions.

• The KIPS Transactions:PartA
• /
• v.12A no.5 s.95
• /
• pp.461-470
• /
• 2005

In leaner-controlled environment where learners can decide and restructure the contents, methods and order of learning by themselves, it is possible to apply individualized learning in consideration of each learner's characteristics. The present study analyzed learners' learning path pattern, which is one of learners' characteristics important in Web-based teaching-learning process, using the Apriori algorithm and grouped learners according to their learning path pattern. Based on the result, we designed and implemented a multi-dimensional learning path pattern analysis system to provide individual learners with teaming paths, learning contents, learning media, supplementary teaming contents, the pattern of material presentation, etc. multi-dimensionally. According to the result of surveying satisfaction with the developed system satisfaction with supplementary learning contents was highest (Highly satisfied '$24.5\%$, Satisfied'$35.7\%$). By learners' level, satisfaction was higher in low-level learners (Highly satisfied'$20.2\%$, Satisfied'$31.2\%$) than in high-level learners (Highly satisfied'$18.4\%$, 'Satisfied'$28.54\%$). The developed system is expected to provide learners with multi-dimensionally meaningful information from various angles using OLAP technologies such as drill-up and drill-down.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1360-1365
• /
• 2019

In this paper, we describe a sound source localization(SSL) system which can be applied to mobile robot and automatic control systems. Usually the SSL method finds the Interaural Time Difference, the Interaural Level Difference, and uses the geometrical principle of microphone array. But here we proposed another approach based on the deep neural network to obtain the horizontal directional angle(azimuth) of the sound source. We pick up the sound source signals from the two microphones attached symmetrically on both sides of the robot to imitate the human ears. Here, we use difference of spectral distributions of sounds obtained from two microphones to train the network. We train the network with the data obtained at the multiples of 10 degrees and test with several data obtained at the random degrees. The result shows quite promising validity of our approach.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.3
• /
• pp.341-348
• /
• 2020

The ability to extract useful information from an image, such as the human eye, is an interface technology essential for AI computer implementation. The building recognition technology has a lower recognition rate than other image recognition technologies due to the various building shapes, the ambient noise images according to the season, and the distortion by angle and distance. The computer vision based building recognition algorithms presented so far has limitations in discernment and expandability due to manual definition of building characteristics. This paper introduces the deep learning CNN (Convolutional Neural Network) model, and proposes new method to improve the recognition rate even by changes of building images caused by season, illumination, angle and perspective. This paper introduces the partial images that characterize the building, such as windows or wall images, and executes the training with whole building images. Experimental results show that the building recognition rate is improved by about 14% compared to the general CNN model.

• Composites Research
• /
• v.34 no.6
• /
• pp.394-399
• /
• 2021

Piecewise Integrated Composite (PIC) is a new concept to design composite structures of multiple stacking angles both for in-plane direction and through the thickness direction in order to improve stiffness and strength. In the present study, PIC beam was suggested based on 3D training data instead of 2D data, which did offer a limited behavior of beam characteristics, with enhancing the stiffness accompanied by reduced tip deformation. Generally training data were observed from the designated reference finite elements, and preliminary FE analysis was conducted with respect to regularly distributed reference elements. Also triaxiality values for each element were obtained in order to categorize the loading state, i.e. tensile, compressive or shear. The main FE analysis was conducted to predict the mechanical characteristics of the PIC beam.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.705-707
• /
• 2022

인체의 흉부 X-ray 영상으로부터 척추질환과 관련된 의료 진단지표를 자동으로 추출하는 과정을 위하여 흉추조직의 정확한 분할이 필요하다. 본 연구에서는 HRNet 기반의 학습을 통하여 흉추조직을 분할하는 방법을 고찰한다. 분할 과정에서 영상 내의 상대적인 위치 정보가 효과적으로 반영될 수 있도록, 계층별로 영상의 고해상도의 표현이 그대로 유지되는 구조와 저해상도의 특징 지도로 변환되는 구조가 병렬적으로 연결되는 형태의 심층 신경망 모델을 채택하였다. 흉부 X-ray 영상에서 콥각도(Cobb's angle)를 산출하는 문제를 대상으로 흉추 분할을 위한 학습 방법, 진단지표 추출 방법 등을 소개하며, 부수적으로 피사체의 위치 변화 및 크기 변화 등에 강인한 성능을 제공하기 위하여 학습 데이터를 증강하는 방법론을 제시하였다. 총 145개의 영상을 사용한 실험을 통하여 제안된 이론의 타당성을 평가하였다.

• Composites Research
• /
• v.35 no.5
• /
• pp.303-308
• /
• 2022

PIC (Piecewise Integrated Composite) is a new concept for designing a composite structure with mosaically assigning various types of stacking sequences in order to improve mechanical properties of laminated composites. Also, machine learning is a sub-category of artificial intelligence, that refers to the process by which computers develop the ability to continuously learn from and make predictions based on data, then make adjustments without further programming. In the present study, the tapered box beam type PIC robot arm for carrying and transferring wide and thin LCD display was designed based on the machine learning in order to increase structural stiffness. Essential training data were collected from the reference elements, which were intentionally designated elements among finite element models, during preliminary FE analysis. Additionally, triaxiality values for each finite element were obtained for judging the dominant external loading type, such as tensile, compressive or shear. Training and evaluating machine learning model were conducted using the training data and loading types of elements were predicted in case the level accuracy was fulfilled. Three types of stacking sequences, which were to be known as robust toward specific loading types, were mosaically assigned to the PIC robot arm. Henceforth, the bending type FE analysis was carried out and its result claimed that the PIC robot arm showed increased stiffness compared to conventional uni-stacking sequence type composite robot arm.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.21 no.5
• /
• pp.595-601
• /
• 2011

This paper presents balancing control of inverted pendulum on the ROBOKER arm using visual information. The angle of the inverted pendulum placed on the robot arm is detected by a stereo camera and the detected angle is used as a feedback and tracking error for the controller. Thus, the overall closed loop forms a visual servoing control task. To improve control performance, neural network is introduced to compensate for uncertainties. The learning algorithm of radial basis function(RBF) network is performed by the digital signal controller which is designed to calculate floating format data and embedded on a field programmable gate array(FPGA) chip. Experimental studies are conducted to confirm the performance of the overall system implementation.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.1
• /
• pp.51-58
• /
• 2023

In this paper, we propose an organ segmentation technique for the automatic extraction of medical diagnostic indicators from X-ray images. In order to calculate diagnostic indicators of heart disease and spinal disease such as VHS(vertebral heart scale) and Cobb angle, it is necessary to accurately segment the thoracic spine, carina, and heart in a chest X-ray image. A deep neural network model in which the high-resolution representation of the image for each layer and the structure converted into a low-resolution feature map are connected in parallel was adopted. This structure enables the relative position information in the image to be effectively reflected in the segmentation process. It is shown that learning performance can be improved by combining the OCR module, in which pixel information and object information are mutually interacted in a multi-step process, and the channel attention module, which allows each channel of the network to be reflected as different weight values. In addition, a method of augmenting learning data is presented in order to provide robust performance against changes in the position, shape, and size of the subject in the X-ray image. The effectiveness of the proposed theory was evaluated through an experiment using 145 human chest X-ray images and 118 animal X-ray images.