• 사물인터넷융복합논문지
• /
• 제4권2호
• /
• pp.1-6
• /
• 2018

이미지와 같은 고차원 데이터로부터 새로운 샘플 데이터를 생성하는 기술은 음성 합성, 이미지 변환 및 이미지 복원 등에 다양하게 활용되고 있다. 본 논문은 고해상도의 이미지들을 생성하는 것과 생성한 이미지들의 variation을 높이기 위한 방안으로 Progressive Growing of Generative Adversarial Networks(PG-GANs)을 구현 모델로 채택하였고, 이를 패션 이미지 데이터에 적용하였다. PG-GANs은 생성자(Generator)와 판별자(discriminator)를 동시에 점진적으로 학습하도록 하는데, 저해상도의 이미지에서부터 계속해서 새로운 레이어들을 추가하여 결과적으로 고해상도의 이미지를 생성할 수 있게끔 하는 방식이다. 또한 생성 데이터의 다양성을 높이기 위하여 미니배치 표준편차 방법을 제안하였고 GAN 모델을 평가하기 위한 기존의 MS-SSIM이 아닌 Sliced Wasserstein Distance(SWD) 평가 방법을 제안하였다.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제10권1호
• /
• pp.9-18
• /
• 2021

생성적 적대 신경망(Generative Adversarial Networks, GANs)은 컴퓨터 비전 분야와 관련 분야에서 큰 인기를 얻었으나, 아직까지는 오디오 신호를 직접적으로 생성하는 GAN이 제시되지 못했다. 오디오 신호는 이미지와 다르게 이산 값으로 구성된 생플링된 신호이므로, 이미지 생성에 널리 사용되는 CNN 구조로 학습하기 어렵다. 이러한 제약을 해결하고자, 최근 GAN 연구자들은 오디오 신호의 시간-주파수 표현을 기존 이미지 생성 GAN에 적용하는 전략을 제안했다. 본 논문은 이 전략을 따르면서 GAN을 사용해 생성된 오디오 신호의 충실도를 높이기 위한 개선된 방법을 제안한다. 본 방법은 공개된 스피치 데이터세트를 사용해 검증했으며, 프레쳇 인셉션 거리(Fréchet Inception Distance, FID)를 사용해 평가했다. 기존의 최신(state-of-the-art) 방법은 11.973의 FID를, 본 연구에서 제안하는 방법은 10.504의 FID를 보였다(FID가 낮을수록 충실도는 높다).

• International Journal of Contents
• /
• 제16권2호
• /
• pp.8-18
• /
• 2020

Data is critical in deep learning but the scarcity of data often occurs in research, especially in the preparation of the paired training data. In this paper, document image binarization with unpaired data is studied by introducing adversarial learning, excluding the need for supervised or labeled datasets. However, the simple extension of the previous unpaired training to binarization inevitably leads to poor performance compared to paired data training. Thus, a new deep learning approach is proposed by introducing a multi-diversity of higher quality generated images. In this paper, a two-stage model is proposed that comprises the generative adversarial network (GAN) followed by the U-net network. In the first stage, the GAN uses the unpaired image data to create paired image data. With the second stage, the generated paired image data are passed through the U-net network for binarization. Thus, the trained U-net becomes the binarization model during the testing. The proposed model has been evaluated over the publicly available DIBCO dataset and it outperforms other techniques on unpaired training data. The paper shows the potential of using unpaired data for binarization, for the first time in the literature, which can be further improved to replace paired data training for binarization in the future.

• International Journal of Internet, Broadcasting and Communication
• /
• 제11권4호
• /
• pp.31-36
• /
• 2019

Unsupervised neural networks have not caught enough attention until Generative Adversarial Network (GAN) was proposed. By using both the generator and discriminator networks, GAN can extract the main characteristic of the original dataset and produce new data with similarlatent statistics. However, researchers understand fully that training GAN is not easy because of its unstable condition. The discriminator usually performs too good when helping the generator to learn statistics of the training datasets. Thus, the generated data is not compelling. Various research have focused on how to improve the stability and classification accuracy of GAN. However, few studies delve into how to improve the training efficiency and to save training time. In this paper, we propose a novel optimizer, named FAST-ADAM, which integrates the Lookahead to ADAM optimizer to train the generator of a semi-supervised generative adversarial network (SSGAN). We experiment to assess the feasibility and performance of our optimizer using Canadian Institute For Advanced Research - 10 (CIFAR-10) benchmark dataset. From the experiment results, we show that FAST-ADAM can help the generator to reach convergence faster than the original ADAM while maintaining comparable training accuracy results.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 춘계학술발표대회
• /
• pp.497-500
• /
• 2022

High-quality images and videos are being generated as technologies for deep learning-based image style translation and conversion of static images into dynamic images have developed. However, it takes a lot of time and resources to manually transform images, as well as professional knowledge due to the difficulty of natural image transformation. Therefore, in this paper, we study natural style mixing through a style conversion network using GAN and natural dynamic image generation using the First Order Motion Model network (FOMM).

• 한국멀티미디어학회논문지
• /
• 제22권5호
• /
• pp.547-557
• /
• 2019

To extract the depth map from a single image, a number of CNN-based deep learning methods have been performed in recent research. In this study, the GAN structure of Pix2Pix is maintained. this model allows to converge well, because it has the structure of the generator and the discriminator. But the convolution in this model takes a long time to compute. So we change the convolution form in the generator to a depthwise convolution to improve the speed while preserving the result. Thus, the seven down-sizing convolutional hidden layers in the generator U-Net are changed to depthwise convolution. This type of convolution decreases the number of parameters, and also speeds up computation time. The proposed model shows similar depth map prediction results as in the case of the existing structure, and the computation time in case of a inference is decreased by 64%.

• 대한한의진단학회지
• /
• 제9권2호
• /
• pp.1-9
• /
• 2005

Background: Liu-Yuan-Lei(陸淵雷) said that a medical record is both the marks of treatments and arts made by a excellent practitioner and the essence of TCM(Traditional Chinese Medicine). Jiang-Guan(江瓘) also said that reading medical records is one of the best way to develop one’s abilities If curing a disease without perfect clinical practice. Objectives: study on the special treatment about metrostaxis(崩漏) based on the Yi-Da-Gan(易大艮)’s medical records. and study on the differentiation of abnormal symptoms and signs about cold syndrome with pesudo-heat(眞寒假熱) based on the Yu-Chang(喩昌)'s medical records. Methods: First, read and study the medical records on metrostaxis(崩漏) of that made a profound study by Yi-Da-Gan(易大艮) and cold syndrome with pesudo-heat(眞寒假熱) of that made a profound study by Yu-Chang(喩昌). The next, write a paper on results and conclusions. Results and Conclusions: First, Yi-Da-Gan(易大艮) insist that must control the Qi under the blood disease conditions, taking the case of metrostaxis(崩漏). Secondly, we must study more on estimating the changing condition of Qi and the blood as time goes by, also study on the pulse and pulse condition in the four seasons(四時脈). Thirdly, Yu-Chang(喩昌) insist that be more careful in differentiation of symptoms and signs, taking the case of cold syndrome with pesudo-heat(眞寒假熱). Fourthly, Yu-Chang(喩昌) give an example that in condition of cold syndrome with pesudo-heat(眞寒假熱), sometimes, the pulse and pulse condition can be strong.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제8권6호
• /
• pp.257-264
• /
• 2019

최근 자연언어처리 분야에서 딥러닝 모델이 좋은 성과를 보이고 있다. 이러한 딥러닝 모델의 성능을 향상시키기 위해서는 많은 양의 데이터가 필요하다. 하지만 많은 양의 데이터를 모으기 위해서는 많은 인력과 시간이 소요되기 때문에 데이터 확장을 통해 이와 같은 문제를 해소할 수 있다. 그러나 문장 데이터의 경우 이미지 데이터에 비해 데이터 변형이 어렵기 때문에 다양한 문장을 생성할 수 있는 생성 모델을 통해 문장 데이터 자동 확장을 해보고자 한다. 본 연구에서는 최근 이미지 생성 모델에서 좋은 성능을 보이고 있는 생성적 적대 신경망 중 하나인 CS-GAN을 사용하여 학습 데이터로부터 새로운 문장들을 생성해 보고 유용성을 다양한 지표로 평가하였다. 평가 결과 CS-GAN이 기존의 언어 모델을 사용할 때보다 다양한 문장을 생성할 수 있었고 생성된 문장을 감성 분류기에 학습시켰을 때 감성 분류기의 성능이 향상됨을 보였다.

• 한국빅데이터학회지
• /
• 제5권2호
• /
• pp.69-76
• /
• 2020

이 연구는 2019년도 정부(과학기술정보통신부)의 재원으로 정보통신기술진흥센터의 지원을 받아 수행된 연구이다. 고병원성조류인플루엔자(Highly Pathogenic Avian Influenza, HPAI)는 병원성이 높은 조류인플루엔자 바이러스 감염에 의하여 발생하는 조류의 급성 전염병으로 닭, 오리 등 가금류에서 피해가 심각하게 나타난다. 고병원성 조류인플루엔자(HPAI)는 연중으로 발생하기보다는 겨울철에 집중하여 발생되는 양상을 보이며, 특정 기간에는 아예 발생하지 않는 경우가 있다. 이와 같은 HPAI의 특성으로 인해 충분한 양의 실제 데이터가 축적되지 못하는 문제점이 있다. 본 논문 연구에서는 GAN 네트워크를 활용하여 결측치를 포함하고 있는 실제와 유사한 데이터를 생성하였으며 해당 과정을 소개한다. 본 연구 결과는 HPAI가 발생하지 않은 특정 시기에 대하여 실제와 유사한 시뮬레이션 데이터를 생성하여 위험도를 측정하는데 이용될 수 있다.

• 한국의사학회지
• /
• 제26권1호
• /
• pp.87-96
• /
• 2013

This study confirmed that a doctor named Lee, Seok-gan whose name has been widely known but whose real identity has remained unclear, was an active Confucian doctor in the 16th century. In addition, through the newly discovered "Daeyakbu" among his family line, writings, and relics that have been handed down in a family, this study looked into his medical philosophy and medicine culture. The author of "Ieseokgangyeongheombang"(Medical Book by Lee, Seok-gan(李石澗), Seok-gan is the same person as an active famous doctor Lee, Seok-gan(李碩幹, 1509-1574) in the 16th century. Such a fact can be confirmed through "Samuiilheombang", "Sauigyeongheombang" and the newly opened "Ieseokgangyeongheombang". Lee, Seok-gan was born in the 4th ruling year of king Jungjong (1509) and was active as a doctor until the 7th ruling year of king Seonjo(1547); his first name is Jungim with the pen name-Chodang, and he used a doctor name of 'Seokgan.' He was known as a divine doctor, and there have been left lots of anecdotes in relation with Lee, Seok-gan. Legend has it that Seokgan went to China to give treatment to the empress, and a heavenly peach pattern drinking cup and a house, which the emperor bestowed on Seokgan in return for his great services, still have remained up to the present. Usually, Seokgan interacted with Toegye Lee Hwang and his literary persons, and with his excellent medical skills, Seokgan once gave treatment to Toegye at the time of his death free of charge. His medical skills have been handed down in his family, and his descendant Lee, Ui-tae(around 1700) compiled a medical book titled "Gyeongheombangwhipyeon(經驗方彙編)". Out of Lee, Seok-gan's keepsakes which were donated to Sosu museums by his descendant family, 4 sorts of 'Gwabu'(writings of fruit trees) including "Daeyakbu" were discovered. It's rare to find a literary work left by a medical figure like this, so these discoveries have a deep meaning even from a medicine culture level. Particularly, "Daeyakbu" includes the typical "Uigukron". The "Uigukron", which develops its story by contrasting politics with medicine, has a unique writing style as one of the representative explanatory methods of scholars' position during the Joseon Dynasty; in addition, the distinctive feature of "Uigukron" is that it was created in the form of 'Gabu' other than a prose.