• 대한토목학회논문집
• /
• 제39권1호
• /
• pp.147-154
• /
• 2019

유한요소모델 업데이팅은 계측에 의한 구조물의 실제 응답과 가장 가까운 응답을 내는 유한요소모델의 매개변수를 찾는 문제로 정의할 수 있다. 기존 연구에서는 실 구조물과 해석 모델의 응답의 오차를 최소화하는 최적화에 기반 한 방법이 개발되었다. 이 연구에서는 목표 모드 정보로부터 유한요소 모델의 매개변수를 직접 얻을 수 있는 역 고유치 문제를 구성하고 역 고유치 문제를 빠르고 정확하게 풀기 위한 깊은 신경망(Deep Neural Network)을 구성하는 방법을 제안한다. 개발한 방법의 적용 예로서 현수교의 역 고유치 함수를 모사하는 신경망을 이용한 동적 유한요소모델 업데이트를 보인다. 해석 결과 제시한 방법은 매우 높은 정확도로 목표 모드에 대응하는 매개변수를 찾아낼 수 있음을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권12호
• /
• pp.6038-6053
• /
• 2017

In this study, we propose an automatic melody extraction algorithm using deep learning. In this algorithm, feature images, generated using the energy of frequency band, are extracted from polyphonic audio files and a deep learning technique, a convolutional neural network (CNN), is applied on the feature images. In the training data, a short frame of polyphonic music is labeled as a musical note and a classifier based on CNN is learned in order to determine a pitch value of a short frame of audio signal. We want to build a novel structure of melody extraction, thus the proposed algorithm has a simple structure and instead of using various signal processing techniques for melody extraction, we use only a CNN to find a melody from a polyphonic audio. Despite of simple structure, the promising results are obtained in the experiments. Compared with state-of-the-art algorithms, the proposed algorithm did not give the best result, but comparable results were obtained and we believe they could be improved with the appropriate training data. In this paper, melody extraction and the proposed algorithm are introduced first, and the proposed algorithm is then further explained in detail. Finally, we present our experiment and the comparison of results follows.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권3호
• /
• pp.877-893
• /
• 2022

With the development of deep learning, face inpainting has been significantly enhanced in the past few years. Although image inpainting framework integrated with generative adversarial network or attention mechanism enhanced the semantic understanding among facial components, the issues of reconstruction on corrupted regions are still worthy to explore, such as blurred edge structure, excessive smoothness, unreasonable semantic understanding and visual artifacts, etc. To address these issues, we propose a Learnable Structure Knowledge of Fusion Network (LSK-FNet), which learns a prior knowledge by edge generation network for image inpainting. The architecture involves two steps: Firstly, structure information obtained by edge generation network is used as the prior knowledge for face inpainting network. Secondly, both the generated prior knowledge and the incomplete image are fed into the face inpainting network together to get the fusion information. To improve the accuracy of inpainting, both of gated convolution and region normalization are applied in our proposed model. We evaluate our LSK-FNet qualitatively and quantitatively on the CelebA-HQ dataset. The experimental results demonstrate that the edge structure and details of facial images can be improved by using LSK-FNet. Our model surpasses the compared models on L1, PSNR and SSIM metrics. When the masked region is less than 20%, L1 loss reduce by more than 4.3%.

• 수산해양기술연구
• /
• 제58권3호
• /
• pp.272-280
• /
• 2022

The first step is to determine the principal dimensions of the design ship, such as length between perpendiculars, beam, draft and depth when accomplishing the design of a new vessel. To make this process easier, a database with a large amount of existing ship data and a regression analysis technique are needed. Recently, deep learning, a branch of artificial intelligence (AI) has been used in regression analysis. In this paper, deep learning neural networks are used for regression analysis to find the regression function between the input and output data. To find the neural network structure with the highest accuracy, the errors of neural network structures with varying the number of the layers and the nodes are compared. In this paper, Python TensorFlow Keras API and MATLAB Deep Learning Toolbox are used to build deep learning neural networks. Constructed DNN (deep neural networks) makes helpful in determining the principal dimension of the ship and saves much time in the ship design process.

• 한국컴퓨터정보학회논문지
• /
• 제26권6호
• /
• pp.1-8
• /
• 2021

본 논문은 저해상도 적외선영상을 사양이 낮은 임베디드 시스템에서 추론 가능하도록 강화된 VGG 스타일과 Global Average Pooling 조합으로 정확도를 증가시키면서 연산량을 최소화하는 딥러닝 컨볼루션 신경망을 이용한 저해상도 적외선 표적 분류 방법을 제안한다. 제안한 알고리즘은 OKTAL-SE로 생성한 합성영상 클래스 9개 3,723,328개를 분류하였다. 최초 임베디드 추론 가능하도록 파라메터 수가 최소화된 최대풀링 레이어 기준 입력단 8개와 출력단 8개 조합에 비해 강화된 VGG 스타일을 적용한 입력단 4개와 출력단 16개 필터수 조합을 이용하여 연산량은 약 34% 감소시켰으며, 정확도는 약 2.4% 증가시켜 최종 정확도 96.1%을 획득하였다. 추가로 C 코드로 포팅하여 수행시간을 확인하였으며, 줄어든 연산량 만큼 수행 시간이 약 32% 줄어든 것을 확인할 수 있었다.

• 한국정보통신학회논문지
• /
• 제24권6호
• /
• pp.809-817
• /
• 2020

소형위성 전력분배 및 전송모듈의 설계와 개발과정에서 딥러닝 알고리즘으로 동적 전력자원의 안정성을 평가하였다. 안정성 평가에 따른 요구사항은 소형위성 탑재체인 SAR 레이더의 전력분배모듈과 수요모듈의 전력전송기능을 구성하였다. 전력모듈인 PDM을 구성하는 스위칭 전력부품의 성능확인을 위해 동적신경망을 활용하여 신뢰성을 검증하였다. 신뢰성 검증을 위한 딥러닝 적용대상은 소형위성 본체로부터 공급되는 전력에 대한 탑재체의 전력분배기능이다. 이 기능에 대한 성능확인을 위한 모델링 대상은 출력전압변화추이(Slew Rate Control), 전압오류(Voltage Error), 부하특성(Load Power)이다. 이를 위해 첫째, 모델링으로 Coefficient Structure 영역을 정의하고 PCB모듈을 제작하여 안정성과 신뢰성을 비교 평가하였다. 둘째, 딥러닝 알고리즘으로 Levenberg-Marquare기반의 Two-Way NARX신경망 Sigmoid Transfer를 사용하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권3호
• /
• pp.881-895
• /
• 2023

Modern image inpainting techniques based on deep learning have achieved remarkable performance, and more and more people are working on repairing more complex and larger missing areas, although this is still challenging, especially for facial image inpainting. For a face image with a huge missing area, there are very few valid pixels available; however, people have an ability to imagine the complete picture in their mind according to their subjective will. It is important to simulate this capability while maintaining the identity features of the face as much as possible. To achieve this goal, we propose a three-stage network model, which we refer to as the identity and structure feature refinement network (ISFRNet). ISFRNet is based on 1) a pre-trained pSp-styleGAN model that generates an extremely realistic face image with rich structural features; 2) a shallow structured network with a small receptive field; and 3) a modified U-net with two encoders and a decoder, which has a large receptive field. We choose structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), L1 Loss and learned perceptual image patch similarity (LPIPS) to evaluate our model. When the missing region is 20%-40%, the above four metric scores of our model are 28.12, 0.942, 0.015 and 0.090, respectively. When the lost area is between 40% and 60%, the metric scores are 23.31, 0.840, 0.053 and 0.177, respectively. Our inpainting network not only guarantees excellent face identity feature recovery but also exhibits state-of-the-art performance compared to other multi-stage refinement models.

• ETRI Journal
• /
• 제42권4호
• /
• pp.505-517
• /
• 2020

The increasing size and complexity of deep neural networks (DNNs) necessitate the development of efficient high-performance accelerators. An efficient memory structure and operating scheme provide an intuitive solution for high-performance accelerators along with dataflow control. Furthermore, the processing of various neural networks (NNs) requires a flexible memory architecture, programmable control scheme, and automated optimizations. We first propose an efficient architecture with flexibility while operating at a high frequency despite the large memory and PE-array sizes. We then improve the efficiency and usability of our architecture by automating the optimization algorithm. The experimental results show that the architecture increases the data reuse; a diagonal write path improves the performance by 1.44× on average across a wide range of NNs. The automated optimizations significantly enhance the performance from 3.8× to 14.79× and further provide usability. Therefore, automating the optimization as well as designing an efficient architecture is critical to realizing high-performance DNN accelerators.

• 말소리와 음성과학
• /
• 제7권4호
• /
• pp.9-16
• /
• 2015

In this paper, several experiments are performed according to deep neural network (DNN) based features for the performance comparison of speaker verification (SV) systems. To this end, input features for a DNN, such as mel-frequency cepstral coefficient (MFCC), linear-frequency cepstral coefficient (LFCC), and perceptual linear prediction (PLP), are first compared in a view of the SV performance. After that, the effect of a DNN training method and a structure of hidden layers of DNNs on the SV performance is investigated depending on the type of features. The performance of an SV system is then evaluated on the basis of I-vector or probabilistic linear discriminant analysis (PLDA) scoring method. It is shown from SV experiments that a tandem feature of DNN bottleneck feature and MFCC feature gives the best performance when DNNs are configured using a rectangular type of hidden layers and trained with a supervised training method.

• International journal of advanced smart convergence
• /
• 제7권2호
• /
• pp.95-100
• /
• 2018

In this paper, we propose a binarized multi-scale module to accelerate the speed of the pose estimating deep neural network. Recently, deep learning is also used for fine-tuned tasks such as pose estimation. One of the best performing pose estimation methods is based on the usage of two neural networks where one computes the heat maps of the body parts and the other computes the part affinity fields between the body parts. However, the convolution filtering with a large kernel filter takes much time in this model. To accelerate the speed in this model, we propose to change the large kernel filters with binarized multi-scale modules. The large receptive field is captured by the multi-scale structure which also prevents the dropdown of the accuracy in the binarized module. The computation cost and number of parameters becomes small which results in increased speed performance.