• Proceedings of the KSRS Conference
• /
• 2009.03a
• /
• pp.296-300
• /
• 2009

가상현실이나 인터넷 웹지도 서비스와 같이 3차원의 실세계를 시스템 상에 그대로 재현(reconstruction)하기 위해서는 정교하고 세밀한 3차원 도시모델이 필수적이다. 이러한 3차원 도시모델의 자동생성은 원격탐사 및 사진측량 분야에서 많은 연구가 수행되고 있다. 이러한 연구들은 다양한 센서 데이터와 기 구축되어 있는 GIS자료를 이용하여 건물, 도로, 지형 등의 도시모델을 자동으로 생성하고자 한다. 그러나 대부분의 연구에서 추출한 각 기본요소(primitives)-평면패치(planar patches), 에지(edges), 모서리(corners)에 대한 국부적인 정제(refinement)는 수행하였으나, 생성한 건물 모델에 대한 광역적인 조정을 통한 정규화에 대한 연구는 미비한 상태이다. 본 연구에서는 다양한 데이터로부터 생성된 B-rep (boundary representation) 형태의 건물 모델에 대하여 기하학적인 제약요소(constraints)를 이용한 정규화(regularization) 방법론을 제시하고자 한다. 제안하는 방법은 건물의 Domain Knowledge에 기반하여 도출한 건물을 구성하는 기본요소(primitives)간의 인접성, 직교성, 평행성, 교차성 등의 다양한 제약조건을 이용하여 광역적으로 조정한다. 시뮬레이션 데이터에 적용한 결과의 분석을 통해 제안된 정규화 방법을 통해 오차가 포함된 건물모델이 보다 정형화된 형태로 조정되었음을 확인하였다.

• International Journal of Highway Engineering
• /
• v.12 no.2
• /
• pp.115-121
• /
• 2010

To evaluate the engineering characteristics of aggregate base materials, cyclic triaxial, CBR and permeability tests were performed for 15 samples. The CBR values of aggregate base materials have wide range from 32 to 110(average 81) and the amount of swelling in submerged conditions has below 0.04mm. The Modulus of aggregate base materials were significantly affected by volumetric stress, linear volumetric model was best for fitting. The modulus of aggregate base materials were determined within range of 100MPa~600MPa, 80~270 and 0.1~0.6 for model coefficient $k_1$ and $k_2$ respectively. The empirical correlation model was suggested that prediction the modulus from the basic properties obtained from particle size distribution test and compaction test. The coefficient of determination of the proposed correlation model was 0.423 for model coefficient $k_1$, 0.920 for model coefficient $k_2$ and 0.872 for modulus with stress level.

• Korean Journal of Remote Sensing
• /
• v.28 no.5
• /
• pp.549-559
• /
• 2012

This paper presents a comparative analysis on the applicability of a priori statistic information about adjustment models when the surface shape parameters are estimated at an arbitrary point in an elevation data. Although the reliability of the estimates are known to be affected by surface condition and the adjustment models, there has been little research in a systematic and detail way. When the raw data have been taken from a real measurement, its true value cannot be known, however, thus this study used simulation data in order to analyze clearly the applicability of adjustment models. The generation of simulated data was performed by superimposing horizontal, slope, and curve surfaces and adding a certain amount of noise. Comparative analysis was performed by associating the a posteriori estimates with a priori statistics of each adjustment models. The experimental results show the estimation characteristics of adjustment models against varying surface conditions.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.49 no.9
• /
• pp.749-757
• /
• 2021

In transportation missions using quadrotor, the payload may change the model parameters, such as mass, moment of inertia, and center of gravity. Moreover, if position of the payload is constantly changing during flight, the effect can adversely affect the control performances. To handle this issue, we suggest Model Reference Adaptive Control based on Linear Quadratic Regulator(LQR+MRAC) to compensate the uncertainty caused by payload. Firstly, the mathematical modeling with the fixed payload is derived. Second, Linear Quadratic Regulator (LQR) is used to design the reference model and baseline controller. Also, through the Stability method, Adaptive law is derived to estimate the model parameters. To verify the performance of proposed control scheme, we compared LQR and LQR+MRAC in situations where uncertainties exist. And, when the disturbance exist, the classic MRAC and proposed controller is compared to analyze the transient response and robustness.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.570-573
• /
• 2024

본 연구는 대형 언어 모델 (LLM) 시대에 공감적 대화 생성을 위한 감정 인식의 필요성을 확인하고 소형 언어 모델 (SLM)을 통한 미세 조정 학습이 고비용 LLM, 특히 ChatGPT의 대안이 될 수 있는지를 탐구한다. 이를 위해 KoBERT 미세 조정 모델과 ChatGPT를 사용하여 사용자 감정을 인식하고, Polyglot-Ko 미세 조정 모델 및 ChatGPT를 활용하여 공감적 응답을 생성하는 비교 실험을 진행하였다. 실험 결과, KoBERT 기반의 감정 분류기는 ChatGPT의 zero-shot 접근 방식보다 뛰어난 성능을 보였으며, 정확한 감정 분류가 공감적 대화의 질을 개선하는 데 기여함을 확인하였다. 이는 공감적 대화 생성을 위해 감정 인식이 여전히 필요하며, SLM의 미세 조정이 고비용 LLM의 실용적 대체 수단이 될 수 있음을 시사한다.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.2 s.40
• /
• pp.285-294
• /
• 2006

The requirements of Computer Supported Cooperative Work supporting efficient cooperative or collaborative works between multi-users have been increasing in distributed environments. The various technical sections such as group communication technology and distributed processing technology should be provided in Cooperative Work. The replicated development of Cooperative Work applications of a number of common requirements increases development costs enormously and duplicated investment parts. Therefore, systematical development environments are required to develop these common requirements and applications efficiently in Cooperative Work applications development. In this study, we propose the hierarchical role-based coordination model that improves the coordination model of legacy mobile agent to be appropriate in Cooperative Work applications, and design the development environment for Cooperative Work applications based on mobile agent. The proposed hierarchical role-based coordination model provides multi-layered group concepts of mobile agent, and enables implementation of efficient coordination policy per group. Additionally, it supports efficient Cooperative Work application development by role assignment per group unit.

• Spatial Information Research
• /
• v.19 no.2
• /
• pp.29-36
• /
• 2011

This study performed quantitative assessment of the performance of adjustment models by a-priori analysis of the statistics of the surface parameter estimates against modeling factors. Lidar, airborne imagery, and SAR imagery have been used to acquire the earth surface elevation, where the shape properties of the surface need to be determined through neighboring observations around target location. In this study, parameters which are selected to be estimated are elevation, slope, second order coefficient. In this study, several factors which are needed to be specified to compose adjustment models are classified into three types: mathematical functions, kernel sizes, and weighting types. Accordingly, a-priori standard deviations of the parameters are computed for varying adjustment models. Then their corresponding confidence regions for both the standard deviation of the estimate and the estimate itself are calculated in association with probability distributions. Thereafter, the resulting confidence regions are compared to each other against the factors constituting the adjustment models and the quantitative performance of adjustment models are ascertained.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.638-643
• /
• 2023

문서 기반 대화 시스템은 크게 질문으로부터 문서를 검색하는 과정과 응답 텍스트를 생성하는 과정으로 나뉜다. 이러한 대화 시스템의 응답 생성 과정에 디코더 기반 LLM을 사용하기 위해서 사전 학습된 LLM을 미세 조정한다면 많은 메모리, 연산 자원이 소모된다. 본 연구에서는 SVD에 기반한 LLM의 경량화를 시도한다. 사전 학습된 polyglot-ko 모델의 행렬을 SVD로 분해한 뒤, full-fine-tuning 해보고, LoRA를 붙여서 미세 조정 해본 뒤, 원본 모델을 미세 조정한 것과 점수를 비교하고, 정성평가를 수행하여 경량화된 모델의 응답 생성 성능을 평가한다. 문서 기반 대화를 위한 한국어 대화 데이터셋인 KoDoc2Dial에 대하여 평가한다.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.30 no.2
• /
• pp.107-116
• /
• 2012

We investigated the accuracy in three dimensional geo-positioning derived by four high resolution satellite images acquired by two different sensors using the vendor-provided rational polynomial coefficients(RPC) based block adjustment in this research. We used two in-track stereo pairs of GeoEye-1 and WorldView-2 satellite and DGPS surveying data. In this experiment, we analyzed accuracies of RPC block adjustment models of two kinds of homogeneous stereo pairs, four kinds of heterogeneous stereo pairs, three 3 triplet image pairs, and one quadruplet image pair separately. The result shows that the accuracies of the models are nearly same. The accuracy without any GCPs reaches about CEP(90) 2.3m and LEP(90) 2.5m and the accuracy with single GCP is about CEP(90) 0.3m and LEP(90) 0.5m.

• Journal of Korea Society of Industrial Information Systems
• /
• v.28 no.5
• /
• pp.15-30
• /
• 2023

Recently, generative models based on the Transformer architecture, such as ChatGPT, have been gaining significant attention. The Transformer architecture has been applied to various neural network models, including Google's BERT(Bidirectional Encoder Representations from Transformers) sentence generation model. In this paper, a method is proposed to create a text binary classification model for determining whether a comment on Korean movie review is positive or negative. To accomplish this, a pre-trained multilingual BERT sentence generation model is fine-tuned and transfer learned using a new Korean training dataset. To achieve this, a pre-trained BERT-Base model for multilingual sentence generation with 104 languages, 12 layers, 768 hidden, 12 attention heads, and 110M parameters is used. To change the pre-trained BERT-Base model into a text classification model, the input and output layers were fine-tuned, resulting in the creation of a new model with 178 million parameters. Using the fine-tuned model, with a maximum word count of 128, a batch size of 16, and 5 epochs, transfer learning is conducted with 10,000 training data and 5,000 testing data. A text sentiment binary classification model for Korean movie review with an accuracy of 0.9582, a loss of 0.1177, and an F1 score of 0.81 has been created. As a result of performing transfer learning with a dataset five times larger, a model with an accuracy of 0.9562, a loss of 0.1202, and an F1 score of 0.86 has been generated.