• 인터넷정보학회논문지
• /
• 제23권4호
• /
• pp.35-43
• /
• 2022

메이크업은 사람의 외모를 개선하는 가장 보편적인 방법이다. 하지만 메이크업의 스타일이 매우 다양하기 때문에 한 개인이 본인에게 직접 메이크업을 하는 것에는 많은 시간적, 비용적 문제점이 존재한다. 이에 따라 메이크업 자동화에 대한 필요성이 증가하고 있다. 메이크업의 자동화를 위해 메이크업 변환(Makeup Transfer)가 연구되고 있다. 메이크업 변환은 메이크업이 없는 얼굴 영상에 메이크업 스타일을 적용시키는 분야이다. 메이크업 변환은 전통적인 영상 처리 기반의 방법과 딥러닝 기반의 방법으로 나눌 수 있다. 특히 딥러닝 기반의 방법에서는 적대적 생성 신경망을 기반으로 한 연구가 많이 수행되었다. 하지만 두 가지 방법 모두 결과 영상이 부자연스럽거나 메이크업 변환의 결과가 뚜렷하지 않고 번지거나 메이크업 스타일 얼굴 영상의 영향을 많이 받는다는 단점이 있다. 메이크업의 뚜렷한 경계를 표현하고 메이크업 스타일 얼굴 영상에서 받는 영향을 완화시키기 위해 본 연구에서는 메이크업 영역을 분할하고 HoG(Histogram of Gradient)를 사용해 손실 함수를 계산한다. HoG는 영상 내에 존재하는 에지의 크기와 방향성을 통해 영상의 특징을 추출하는 방법이다. 이를 통해 에지에 대해 강건한 학습을 수행하는 메이크업 변환에 대해 제안한다. 제안한 모델을 통해 생성된 영상과 베이스 모델로 사용하는 BeautyGAN을 통해 생성된 영상을 비교해 본 연구에서 제안한 모델의 성능이 더 뛰어남을 확인하고 추가로 제시할 수 있는 얼굴 정보에 대한 사용 방법을 향후 연구로 제시한다.

• Nuclear Engineering and Technology
• /
• 제54권5호
• /
• pp.1890-1901
• /
• 2022

We have assessed the applicability of the thermal-hydraulic system analysis code, SPACE, to a small modular reactor called SMART. For the assessment, the experimental data from a scale-down integral-test facility, SMART-ITL, were used. It was conformed that the SPACE code unrealistically calculates the safety injection flow rate through the CMT and SIT during a small-break loss-of-coolant experiment. This unrealistic behavior was due to the overprediction of interfacial heat transfer at the steam-water interface in a vertically stratified flow in the tanks. In this study, a special thermal-hydraulic component model has been developed to realistically calculate the interfacial heat transfer when a strong non-equilibrium two-phase flow is formed in the CMT or SIT. Additionally, we developed a special heat structure model, which analytically calculates the heat transfer from the hot steam to the cold tank wall. The combination of two models for the tank are called the special component model. We assessed it using the SMART-ITL passive safety injection system (PSIS) test data. The results showed that the special component model well predicts the transient behaviors of the CMT and SIT.

• 한국원자력학회:학술대회논문집
• /
• 한국원자력학회 1996년도 춘계학술발표회논문집(2)
• /
• pp.683-690
• /
• 1996

The MIDLOOP code has been developed for the evaluation of RES pressurization transients initiated from a loss-of-Residual Heat Removal System (RHRS) during mid-loop operation after reactor shutdown. It provides a fast running and realistic tool for studying parametrically the response of important plant parameters such as pressure, temperature, and level to various plant combinations of the primary side vent, makeup, and leakage procedures and the steam generator (SG) conditions. The code consists of ten nodes representing the primary and secondary sides of a nuclear power plant and can analyze the effect of air on the primary system pressurization and primary to secondary heat transfer. The analysis results of the MIDLOOP code are in good agreement with the ROSA-IV/LSTF experiment without opening in the RCS.

• 한국원자력학회:학술대회논문집
• /
• 한국원자력학회 1996년도 춘계학술발표회논문집(2)
• /
• pp.368-373
• /
• 1996

A simple set of the transition criterion of the condensation regimes and the heat transfer coefficients on the direct contact condensation of the core makeup tank is developed, and implemented in RELAP5/MOD3.1 The condensation regimes are divided into two regimes: supply limit and condensation limit. In mode]ing the transition criterion between two regimes, a large-eddy model developed by Theofanous is used, and the empirical coefficient of the present large-eddy model is close to that of the large-eddy model. It turns out that the modified code better predicts the experimental data, especially the injection flow rate and the water level trend than the original code does.

• International Journal of Computer Science & Network Security
• /
• 제21권8호
• /
• pp.127-136
• /
• 2021

Real-time human recognition is a challenging task, as the images are captured in an unconstrained environment with different poses, makeups, and styles. This limitation is addressed by generating several facial images with poses, makeup, and styles with a single reference image of a person using Generative Adversarial Networks (GAN). In this paper, we propose deep learning-based human recognition using integration of GAN and Spatial Domain Techniques. A novel concept of human recognition based on face depiction approach by generating several dissimilar face images from single reference face image using Domain Transfer Generative Adversarial Networks (DT-GAN) combined with feature extraction techniques such as Local Binary Pattern (LBP) and Histogram is deliberated. The Euclidean Distance (ED) is used in the matching section for comparison of features to test the performance of the method. A database of millions of people with a single reference face image per person, instead of multiple reference face images, is created and saved on the centralized server, which helps to reduce memory load on the centralized server. It is noticed that the recognition accuracy is 100% for smaller size datasets and a little less accuracy for larger size datasets and also, results are compared with present methods to show the superiority of proposed method.

• 한국응용과학기술학회지
• /
• 제34권4호
• /
• pp.727-736
• /
• 2017

본 연구에서는 임상시험의 대체시험법으로 시편을 이용한 화장품의 보습능 평가기법을 연구하였다. 4종의 시편(일본떡, 가래떡, 밀가루, 한천)에 대한 립 메이크업 제품의 발림성, 표면의 건조 특성, 수분손실율 등을 비교 평가하였다. 또한, 10종의 립 메이크업 제품(립스틱 5종, 립밤 3종, 립글로스 2종)을 대상으로 시편을 사용한 수분손실율과 임상평가를 통한 경피수분손실량을 분석하고 이들의 상관 관계를 조사하였다. 시편의 수분손실율은 한천 > 가래떡 > 일본떡 > 밀가루 순으로 한천시편의 수분 손실율(수분증발 민감도)이 가장 우수하였다. 또한, 시편소재 중 한천시편이 발림성, 표면 균열, 원료 수급 등의 측면에서 가장 적합하였다. 한천시편은 볼록하고 매끄러운 표면의 시편 제조가 가능한 열 전달이 낮은 플라스틱재질의 용기가 적합하였다. 한천시편을 이용한 립메이크업 제품군에 대한 수분손실율(보습능)과 임상시험법(경피수분손실)과 강한 상관관계를 보였다. 이러한 결과들은 본 연구에서 제안한 한천 시편을 이용한 보습능 평가 기법이 유용한 대체시험법 중 하나로 활용 가능함을 의미한다.

• 한국토양비료학회지
• /
• 제46권2호
• /
• pp.73-80
• /
• 2013

Due to the fact that possible risk associated with soil-crop-food chain transfer, metal contamination in croplands has become a major topic of wide concern. Accumulation of toxic metals in edible parts of crops grown in contaminated soils has been reported from number of crops including rice, soybean, wheat, maize, and vegetables. Therefore, in order to ensure food safety, measures are needed to be taken in mitigating metal pollution and subsequent uptake by crop plants. Present paper critically reviewed some of the cost effective remediation techniques used in minimizing metal uptake by crops grown in contaminated soils. Liming with different materials such as limestone ($CaCO_3$), burnt lime (CaO), slaked lime [$Ca(OH)_2$], dolomite [$CaMg(CO_3)_2$], and slag ($CaSiO_3$) has been widely used because they could elevate soil pH rendering metals less-bioavailable for plant uptake. Zn fertilization, use of organic amendments, crop rotation and water management are among the other techniques successfully employed in reducing metal uptake by crop plants. However, irrespectively the mitigating measure used, heterogeneous accumulation of metals in different crop species is often reported. The inconsistency might be attributed to the genetic makeup of the crops for selective uptake, their morphological characteristics, position of edible parts on the plants in respect of their distance from roots, crop management practices, the season and to the soil characteristics. However, a sound conclusion in this regard can only be made when more scientific evidence is available on case-specific researches, in particular from long-term field trials which included risks and benefits analysis also for various remediation practices.