• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.221-223
• /
• 2022

최근 CNN(Convolutional Neural Network)은 초해상화(super-resolution)를 포함한 다양한 컴퓨터 비전 분야에서 우수한 성능을 보이며 널리 사용되고 있다. 그러나 CNN은 계산 집약적이고 많은 메모리가 요구되어 한정적인 하드웨어 자원인 모바일이나 IoT(Internet of Things) 기기에 적용하기 어렵다는 문제가 있다. 이런 한계를 해결하기 위해, 기 학습된 깊은 CNN 모델의 성능을 최대한 유지하며 네트워크의 깊이나 크기를 줄이는 경량화 연구가 활발히 진행되고 있다. 본 논문은 네트워크 경량화 기술인 지식증류(knowledge distillation) 중 자가증류(self-distillation)를 초해상화 CNN 모델에 적용하여 성능을 평가, 분석한다. 실험 결과, 정량적 평가지표를 통하여 자가증류를 통해서도 성능이 우수한 경량화된 초해상화 모델을 얻을 수 있음을 확인하였다.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 2017.04a
• /
• pp.166-166
• /
• 2017

모돈의 건강 상태를 정량 지수화 하기 위한 연구를 수행 중이다. 지제이상, 섭식 불량, 수면 패턴 등의 운동 특성 분석을 위하여 복수의 초음파 센서를 이용하였다. 시계열 계측 신호를 분석하여 정량 지수화를 수행하는 과정에서 주파수 도메인 분석을 시도하였다. 이 과정에서 주파수 도메인의 분해능에 따른 편차 극복을 위한 비선형 모델링을 수행하였다. 또한 인접한 시계열 데이터 구간 간의 상관성 분석이 가능하면 대용량 데이터의 실시간 처리로 인한 지연 시간 극복 및 기대되는 예후에 대한 조기 진단이 가능할 것이다. 본 연구에서는 구글에서 제공하는 Tensorflow와 NVIDIA에서 제공하는 CUDA 엔진을 동시 적용한 심층 학습 시스템을 이용하였다. 전 처리를 위하여 주파수 분해능 (2분, 3분, 5분, 7분, 11분, 13분, 17분, 19분)에 따른 데이터 집합을 1단계로 두고, 상위 10 순위 안에 드는 파워 스펙트럼 밀도의 크기를 2단계로 하여, 총 2~10개의 입력 노드를 순차적으로 선정하였고, 동일한 방식으로 인접한 시계열의 파워 스펙터럼 밀도를 순위를 변화시켜 지정하였다. 대표적인 심층학습 모델인 Softmax regression with a multilayer convolutional network를 이용하여 Recursive feature selection 경우의 수를 $8{\times}9{\times}9$로 총 648 가지 선정하고, Epoch는 10,000회로 지정하였다. Calibration 모델링의 경우 Cost function이 10% 이하인 경우 해당 경우의 학습을 중단하였으며, 모델 간 상호 교차 검증을 수행하기 위하여 $_8C_2{\times}_8C_2{\times}_8C_2$ 경우의 수에 대한 Verification test를 수행하였다. Calibration 과정 상 모든 경우에 대하여 10% 이하의 Cost function 값을 보였으나, 검증 테스트 과정에서 모든 경우에 대하여 $r^2$ < 0.5 인 결정 계수 값이 나타났다. 단적으로 심층학습 모델의 과도한 적합(Over fitting) 방식의 한계를 보인 것이라 판단할 수 있다. 적합한 Feature selection 및 심층 학습 모델에 대한 지속적이고 추가적인 고려를 통해 과도적합을 해소함과 동시에 실효적이고 활용 가능한 Classification을 위한 입, 출력 노드 단의 전후 Indexing, Quantization에 대한 고려가 필요할 것이다. 이를 통해 모돈 생체 정보 정량화를 위한 지능형 현장 진단 기술 연구를 지속할 것이다.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.7 no.4
• /
• pp.269-283
• /
• 2005

To figure out the cause of underestimating the roughness and shear strength of rock joints suggested by numerous researchers, we analyzed roughness mobilization characteristics, characteristics of roughness parameters, effects of sampling interval, and waviness for roughness parameters. It was found out that lack of understanding of the roughness mobilization characteristics, inappropriate applications of roughness parameters, and effect of aliasing provide a main reasons for those problems. Several practical alternatives for improving those problems were suggested. As far as digitizing methods are concerned, we can find that using a 3D scanner can give a relatively effective result. To avoid aliasing, sampling interval should be less than one-quarter of the minimum asperities. As for the quantification of roughness, it was analyzed that the roughness parameter should be classified into two components depending on the scale of roughness to apply the shear strength model. For classifying the roughness, a framework of the criterion was suggested based on the plastic flow concept for the asperity failure, and the basis for proposing a new alternative shear strength model was established.

• Korean Journal of Soil Science and Fertilizer
• /
• v.41 no.4
• /
• pp.267-272
• /
• 2008

Mid-infrared (MIR) spectroscopy, particularly Fourier transform infrared spectroscopy (FTIR), has emerged as an important analytical tool in quantification as well as identification of multi-atomic inorganic ions such as nitrate. In the present study, the possibility of quantifying soil nitrate via diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) without change of a sample phase or with least treated samples was examined. Four types of soils were spectrally characterized in terms of unique bands of soil contents and interferences with nitrate bands in the range of $2000-1000cm^{-1}$. In order to reduce the effects of soil composition on calibration model for nitrate, spectra transformed to the 1st order derivatives were used in the partial least squared regression (PLSR) model and the classification procedure associated with input soil types was involved in calibration system. PLSR calibration models for each soil type provided better performance results ($R^2$>0.95, RPD>6.0) than the model considering just one type of soil as a standard.

• Nuclear Engineering and Technology
• /
• v.26 no.1
• /
• pp.1-8
• /
• 1994

The flame speed correlation considering thermal-hydraulic phenomena under severe accidents is proposed and correction coefficients are defined. This correlation modifies the pressure dependency in Iijima-Takeno correlation and adds the steam suppression effects to it in the anticipated hydrogen and steam concentration ranges under severe accidents. The existing models of flame speed due to hydrogen combustion under severe accidents are based on the experiments which were performed merely at room temperature and atmospheric pressure. They have difficulty in predicting a accurate flame speed in a case of high temperature and pressure during severe accidents. Thus the flame structure is assumed as a prerequisite to the reliable determination of flame speed and theoretical model is developed. To examine the validity, flame speeds in various conditions calculated by this model are compared with those obtained by the calculation of the existing correlations of the codes such as improved HECTR and MAAP. Also the steam suppression ratio is quantified and the steam suppression coefficient is defined as a composition of mixture. Initial temperature and pressure dependencies are investigated and correction coefficents are determined. More experimental studies can be recommended to improve this correlation to its further works.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.193-194
• /
• 2021

본 논문에서는 지식 증류 (knowledge distillation) 기법을 적용한 얼굴 초해상화 모델을 제안한다. 제안하는 기법은 최근 얼굴 복원 분야에서 좋은 성능을 보여준 얼굴 영역의 딕셔너리 (dictionary) 정보를 사용한 모델을 선생 모델로 선정하여 적대적 (adversarial) 지식 증류 기법을 통해 효율적인 학생 모델을 구축하였다. 본 논문은 테스트시 얼굴의 사전 정보가 초래하는 추가적인 비용이 필요 없는 얼굴 초해상화 방법을 제시하고, 제안하는 기법과 다양한 기존 초해상화 기법과의 정량적, 정성적 비교를 통해 우수성을 보인다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.10b
• /
• pp.919-922
• /
• 2000

본 논문에서는 가상 인터페이스로서 범용적으로 사용할 수 있는 실시간 비전 기반 얼굴 애니메이션을 제안한다. 이를 위해서 실시간으로 얼굴 구성요소를 추출하고, 이를 정량화 하였다. 정량화된 값과 자연스럽게 매핑하기 위해서 정합함수를 통해 3차원 모델에 맵핑하기 위한 방안을 제안한다. 일반적으로 3차원 애니메이션을 수행할 경우, 기본 모델을 중심으로 특정한 사용자만을 위주로 수행되나, 본 논문에서는 임의의 일반 사용자를 위한 3차원 애니메이션을 수행하였다. 여러 사용자에 대하여 얼굴 구성요소 추출을 이용한 3차원 얼굴 애니메이션 동작에 대하여 실험하였으며 실험결과 여러 사용자 얼굴 애니메이션 동작에 대하여 만족할 만한 성능을 보였다.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2022.04a
• /
• pp.206-207
• /
• 2022

Currently, safety accidents in construction area are managed regardless of the size. Therefore, the objective of this study is to conduct for developing the quantitative risk assessment according to large and small and medium-sized construction sites. The scope of this study is limited to the fall accidents which is the biggest accidents in the construction sites. the regression analysis was conducted based on the collected data. As a result, it was confirmed that there was a statistically significant difference between larce and small and medium-sized construction sites. This study is expected to be used as basic data for research on the development of a risk quantitative model for small and medium-sized construction sites in the future.

• Journal of the Korean Society of Radiology
• /
• v.16 no.3
• /
• pp.295-302
• /
• 2022

Sparse-view computed tomography (CT) imaging technique is able to reduce radiation dose, ensure the uniformity of image characteristics among projections and suppress noise. However, the reconstructed images obtained by the sparse-view CT imaging technique suffer from severe artifacts, resulting in the distortion of image quality and internal structures. In this study, we proposed a convolutional neural network (CNN) with wavelet transformation and residual learning for reducing artifacts in sparse-view CT image, and the performance of the trained model was quantitatively analyzed. The CNN consisted of wavelet transformation, convolutional and inverse wavelet transformation layers, and input and output images were configured as sparse-view CT images and residual images, respectively. For training the CNN, the loss function was calculated by using mean squared error (MSE), and the Adam function was used as an optimizer. Result images were obtained by subtracting the residual images, which were predicted by the trained model, from sparse-view CT images. The quantitative accuracy of the result images were measured in terms of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). The results showed that the trained model is able to improve the spatial resolution of the result images as well as reduce artifacts in sparse-view CT images effectively. Also, the trained model increased the PSNR and SSIM by 8.18% and 19.71% in comparison to the imaging model trained without wavelet transformation and residual learning, respectively. Therefore, the imaging model proposed in this study can restore the image quality of sparse-view CT image by reducing artifacts, improving spatial resolution and quantitative accuracy.

• Journal of the Korean Chemical Society
• /
• v.68 no.2
• /
• pp.107-116
• /
• 2024

The purpose of this study is to analyze changes in pre-service chemistry teachers' cognition of the nature of model in the evaluation and modification process of model using technology. Changes in cognition of the nature of model were analyzed focusing on the 'Abstraction' and 'Simplification' of the 'Representational aspect', 'Interpretation', 'Reasoning', 'Explanation' and 'Quantification' of the 'Explanatory aspect' that were deemed insufficient for pre-chemistry teachers in previous study. For this purpose, 19 third-year pre-service chemistry teachers at a teacher's college in Chungcheongbuk-do were asked to evaluate the model related to Boyle's law developed using technology, revise the model based on the evaluation results, and make a final evaluation. As a result of the study, it was confirmed that pre-service chemistry teachers' cognition of 'Simplification' of the 'Representational aspect' and 'Interpretation', 'Explanation', and 'Quantification' of the 'Explanatory aspect' changed positively through the evaluation and modification process of the model. Therefore, it was found that the evaluation and modification process of the model plays a key role in changing the cognition of the nature of model. However, there was little change in cognition of 'Abstraction' of the 'Representational aspect' and 'Reasoning' of the 'Explanatory aspect'. The cognition of these factors can be seen as more difficult to change than the cognition of other factors. To solve this problem, more sophisticated educational design for pre-service chemistry teachers is needed.