• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2021년도 추계학술발표대회
• /
• pp.764-767
• /
• 2021

While Generative Adversarial Networks (GANs) have seen huge success in image synthesis tasks, synthesizing high-quality images from text descriptions is a challenging problem in computer vision. This paper proposes a method named Text-to-Face Generation Using Multi-Scale Gradients for Conditional Generative Adversarial Networks (T2F-MSGGANs) that combines GANs and a natural language processing model to create human faces has features found in the input text. The proposed method addresses two problems of GANs: model collapse and training instability by investigating how gradients at multiple scales can be used to generate high-resolution images. We show that T2F-MSGGANs converge stably and generate good-quality images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권5호
• /
• pp.1431-1445
• /
• 2022

We construct a deep neural network model named ECGResNet. This model can diagnosis diseases based on 12-lead ECG data of eight common cardiovascular diseases with a high accuracy. We chose the 16 Blocks of ResNet50 as the main body of the model and added the Squeeze-and-Excitation module to learn the data information between channels adaptively. We modified the first convolutional layer of ResNet50 which has a convolutional kernel of 7 to a superposition of convolutional kernels of 8 and 16 as our feature extraction method. This way allows the model to focus on the overall trend of the ECG signal while also noticing subtle changes. The model further improves the accuracy of cardiovascular and cerebrovascular disease classification by using a fully connected layer that integrates factors such as gender and age. The ECGResNet model adds Dropout layers to both the residual block and SE module of ResNet50, further avoiding the phenomenon of model overfitting. The model was eventually trained using a five-fold cross-validation and Flooding training method, with an accuracy of 95% on the test set and an F1-score of 0.841.We design a new deep neural network, innovate a multi-scale feature extraction method, and apply the SE module to extract features of ECG data.

• Structural Engineering and Mechanics
• /
• 제53권3호
• /
• pp.393-410
• /
• 2015

Two dimensional numerical models and physical models have been developed to study the highly nonlinear interactions between waves and breakwaters, but several of these models consider the effects of the structural dynamic responses and the shape of the breakwater axis on the wave pressures. In this study, a multi-material Arbitrary Lagrangian Eulerian (ALE) method is developed to simulate the nonlinear interactions between nonlinear waves and elastic seawalls on a coastal rubble mound breakwater, and is validated experimentally. In the experiment, a solitary wave is generated and used with a physical breakwater model. The wave impact is validated computationally using a breakwater - flume coupling model that replicates the physical model. The computational results, including those for the wave pressure and the water-on-deck, are in good agreement with the experimental results. A local breakwater model is used to discuss the effects of the structural dynamic response and different design parameters of the breakwater on wave loads, together with pressure distribution up the seawall. A large-scale breakwater model is used to numerically study the large-scale wave impact problem and the horizontal distribution of the wave pressures on the seawalls.

• 유통과학연구
• /
• 제12권9호
• /
• pp.95-99
• /
• 2014

Purpose - This study aims to develop and empirically test a multi-dimensional service convenience scale with the dimensions and measurement items of service convenience perceived by Korean consumers in retail contexts. Research design, data, and methodology - The study adopts the framework suggested by Berry et al. (2002) and conceptualizes service convenience as a second-order reflective construct comprising 31 items. Three department store chains (Hyundai, Lotte, and Shinsegae) and three discount store chains (E-mart, Homeplus, and Lotte Mart) were involved 510 valid responses were used for the empirical testing. Results - The measurement model is acceptable for internal consistency, convergent validity, and discriminant validity. Further, the structural model results show that service convenience is positively related to satisfaction. Results of the rival model comparison indicate that the proposed second-order factor model provides a better fit to the data than both the five-factor and the one-factor model. Conclusions - The multi-dimensional, second-order conceptualization of service convenience is robustly supported. This study provides psychometrically valid scales to measure service convenience in retail contexts as conceptualized by Berry et al. (2002).

• Nuclear Engineering and Technology
• /
• 제55권9호
• /
• pp.3367-3382
• /
• 2023

Smoothed Particle Hydrodynamics (SPH) is a Lagrangian computational fluid dynamics method that has been widely used in the analysis of physical phenomena characterized by large deformation or multi-phase flow analysis, including free surface. Despite the recent implementation of eddy-viscosity models in SPH methodology, sophisticated turbulent analysis using Lagrangian methodology has been limited due to the lack of computational performance and numerical consistency. In this study, we implement the standard and dynamic Smagorinsky model and dynamic Vreman model as sub-particle scale models based on a weakly compressible SPH solver. The large eddy simulation method is numerically identical to the spatial discretization method of smoothed particle dynamics, enabling the intuitive implementation of the turbulence model. Furthermore, there is no additional filtering process required for physical variables since the sub-grid scale filtering is inherently processed in the kernel interpolation. We simulate lid-driven flow under transition and turbulent conditions as a benchmark. The simulation results show that the dynamic Vreman model produces consistent results with experimental and numerical research regarding Reynolds averaged physical quantities and flow structure. Spectral analysis also confirms that it is possible to analyze turbulent eddies with a smaller length scale using the dynamic Vreman model with the same particle size.

• 전력전자학회논문지
• /
• 제19권1호
• /
• pp.31-40
• /
• 2014

This paper describes an optimal configuration for multi-central inverters in a medium-voltage (MV) grid, which is suitable for large-scale photovoltaic (PV) power plants. We theoretically analyze a proposed common-mode equivalent model for problems associated with multi-central transformerless-type three-phase full bridge(3-FB) PV inverters employing two-winding MV transformers. We propose a synchronized PWM control strategy to effectively reduce the common-mode voltages that may simultaneously occur. In addition, we propose that the existing 3-FB topology may also have the configuration of a multi-central inverter with a two-winding MV transformer by making a simple circuit modification. Simulation and experimental results of three 350kW PV inverters in a multi-central configuration verify the effectiveness of the proposed synchronization control strategy. The modified transformerless-type 3-FB topology for a multi-central PV inverter configuration is verified using an experimental prototype of a 100kW PV inverter.

• 한국음향학회:학술대회논문집
• /
• 한국음향학회 1991년도 학술발표회 논문집
• /
• pp.58-63
• /
• 1991

This paper describes development of small scale voice dialing system using TMS320C30. Recognition vocabuliary is used 50 department name within university. In vocabulary below the middle scale, word unit recognition is more practice than phoneme unit or syllable unit recognition. In this paper, we performend recognition and model generation using DMS(Dynamic Multi-Section) and implemeted voice dialing system using TMS320C30. As a result of recognition, we achieved a 98% recognition rate in condition of section 22 and weight 0.6 and recognition time took 4 seconds.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2004년도 ICCAS
• /
• pp.690-693
• /
• 2004

This paper proposes a multi-resolution electrostatic potential field (MREPF) based solution to the mobile robot path planning and collision avoidance problem in 2D dynamic environment. The MREPF is an environment method in calculation time and updating field map. The large scale resolution map is added to EPF and this resolution map interacts with the small scale resolution map to find an optimal solution in real time. This approach can be interpreted with Atlantis model. The simulation studies show the efficiency of the proposed algorithm.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2006년도 추계학술대회 논문집 전력기술부문
• /
• pp.342-344
• /
• 2006

The modem enormous electric power system has made power system analysis much more complex and difficult. For effective analysis of the power system, model reduction and aggregation is required. In this paper, a new aggregation method is presented to aggregate the coherent generators in the large scale power system while matching the power flow. In order to demonstrate the effects of this aggregation method, it is applied to a small scale power system. A multi-variable control technique of $H_{\infty}$ control is also applied to enhance the dynamic stability of the aggregated power system.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 2007년도 추계학술대회 논문집
• /
• pp.344-347
• /
• 2007

For modeling the non-periodic and randomly scattered powder particles, the quasi-random multi-particle array is introduced. The multi-scale process simulation, which enables to formulate a regression model with a response surface method, is performed by employing a homogenization method. The size of ${Al_2}{O_3}$ particle, amplitude of cyclic compaction pressure, and friction coefficient are considered as optimal process parameters. The optimal conditions of process parameters providing the highest relative density are finally found by using the grid search method.