http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
결절성 폐암 검출을 위한 상용 및 맞춤형 CNN의 성능 비교
박성욱,김승현,임수창,김도연 한국멀티미디어학회 2020 멀티미디어학회논문지 Vol.23 No.6
Screening with low-dose spiral computed tomography (LDCT) has been shown to reduce lung cancer mortality by about 20% when compared to standard chest radiography. One of the problems arising from screening programs is that large amounts of CT image data must be interpreted by radiologists. To solve this problem, automated detection of pulmonary nodules is necessary; however, this is a challenging task because of the high number of false positive results. Here we demonstrate detection of pulmonary nodules using six off-the-shelf convolutional neural network (CNN) models after modification of the input/output layers and end-to-end training based on publicly databases for comparative evaluation. We used the well-known CNN models, LeNet-5, VGG-16, GoogLeNet Inception V3, ResNet-152, DensNet201, and NASNet. Most of the CNN models provided superior results to those of obtained using customized CNN models. It is more desirable to modify the proven off-the-shelf network model than to customize the network model to detect the pulmonary nodules.
전이학습에 방법에 따른 컨벌루션 신경망의영상 분류 성능 비교
박성욱,김도연 한국멀티미디어학회 2018 멀티미디어학회논문지 Vol.21 No.12
Core algorithm of deep learning Convolutional Neural Network(CNN) shows better performance than other machine learning algorithms. However, if there is not sufficient data, CNN can not achieve satisfactory performance even if the classifier is excellent. In this situation, it has been proven that the use of transfer learning can have a great effect. In this paper, we apply two transition learning methods(freezing, retraining) to three CNN models(ResNet-50, Inception-V3, DenseNet-121) and compare and analyze how the classification performance of CNN changes according to the methods. As a result of statistical significance test using various evaluation indicators, ResNet-50, Inception-V3, and DenseNet-121 differed by 1.18 times, 1.09 times, and 1.17 times, respectively. Based on this, we concluded that the retraining method may be more effective than the freezing method in case of transition learning in image classification problem.
박성욱,김일곤,유동선 한국물리학회 2015 새물리 Vol.65 No.6
CdSe/CdS quantum dots (QDs) were synthesized by using different doses from an electron beam (E-B) and several molar ratios of [Se] in the solution. These QDs were characterized by using optical absorption, photoluminescence, X-ray diffraction (XRD), Fourier-transform infrared spectroscopy (FTIR) and high-resolution transmission microscopy (HRTEM) measurements. The absorption spectrum of CdSe/CdS QDs shifts to longer wavelength with increasing dose, as does the photoluminescence (PL) spectrum. The Stokes shift was about 30 nm. This result implies that CdSe/CdS core/shell QDs were synthesized. Under an E-B dose of 40 kGy, a broad spectrum is observed around a wavelength of 600 nm due to surface defects in QDs. This band disappeared for doses above 60 kGy. As the molar ratio of [Se], was increased the absorption and the PL spectra of the QDs shifted toward the red, and the Stokes’ shift decreased linearly. For extracting the QDs from the precipitate synthesized by using the E-B method, we used a new washing solvent that included chloroform and acetonitrile (AN). The washed nanoparticles exhibited PL quantum efficiencies that were four times higher than the intensities of the original nanoparticles. 전자빔의 조사량과 전구물질에서의 [Se]의 몰 비율을 변화시키면서 양자점을 합성하였으며, 합성한 양자점의 물리 및 광학적인 특성을 조사하였다. 전자빔의 조사량이 증가하면 CdSe/CdS 양자점의 흡수스펙트럼과 PL 스펙트럼이 장파장 쪽으로 이동하며, 스토크 이동 (Stokes shifts)은 수십 nm 임을 확인하였다. 이로부터 CdSe/CdS 코어/셀 구조의 양자점이 합성되었음을 알 수 있었다. 전자빔의 세기가 40 kGy 이하에서는 PL 스펙트럼의 600 nm 부근에 표면 결함에 의한 폭넓은 밴드가 있음을 확인하였다. 이 결함은 전자빔의 조사량이 60 kGy 이상이 되면 사라졌다. [Cd]와 [MPA]의 몰 비율을 고정시킨 상태로 [Se] 몰 비율을 증가시키면 흡수스펙트럼과 발광스펙트럼은 적색천이 하며, 이때 스토크 이동량은 선형적으로 감소하였다. 전자빔에 의해 합성된 양자점을 침전물에서 분리하기 위해서 새로운 형태의 세정액 (클로로포름과 아세토나이트릴)을 사용하였으며, 세정한 양자점의 PL 양자효율이 4배 증가함을 확인하였다.
박성욱 대한전자공학회 2006 電子工學會論文誌 IE (Industry electronics) Vol.43 No.1
본 논문에서는 제한된 자원을 가지는 포터블 임베디드 시스템을 위하여, 작은 메모리 사용으로 효율적으로 영상 부호화가 가능한 웨이블릿 영상 부호화기를 제안하였다. 제안된 방법은 부호화 과정시 요구되는 메모리 사용량을 줄이기 위해 웨이블릿 계수들의 비트 레벨 정보를 가지는 2차원 중요 계수 배열을 사용한다. 제안된 방법은 중요 계수에 대한 부호화 과정과 계수들의 비트 레벨 정보의 부호화 과정을 한 번에 수행할 수 있다. 실험 결과 기존의 부호화 방법보다 화질 면에서 비슷하거나 우수한 성능을 보였으며, 2차원 중요 계수 배열을 이용한 최소의 메모리 사용으로 다양한 비트율에서 영상의 일그러짐 없이 안정적으로 동작함을 확인하였다. In order to provide an efficient way to processing with limited resources, we propose a wavelet coder that operates with little memory usage on the portable embedded system. In order to reduce redundancy in coding process caused by repetitive scanning of wavelet coefficients, the proposed coder uses a 2D significance coefficient array (SCA) which records the bit-level information of wavelet coefficients. The 2D SCA improves memory usage and processing speed required for image coding because it can perform significance check and bit coding of coefficients simultaneously
박성욱,박문수,Soo-Hyun Park,Young-Mi Yun 아시아기술혁신학회 2020 Asian Journal of Innovation and Policy Vol.9 No.3
The policy change in the Data 3 Act is one of the issues that should be noted at a time when non-face-to-face business strategies become important after COVID-19. The Data 3 Act was implemented in South Korea on August 5, 2020, calling ‘Big Data 3 Act’ and ‘Data Economy 3 Act,’ and so personal information that was not able to identify a particular individual could be utilized without the consent of the individual. With the implementation of the Data 3 Act, it is possible to establish a fair economic ecosystem by ensuring fair access to data and various uses. In this paper, the law on the protection of personal information, which is the core of the Data 3 Act, was compared around Korea, the European Union and the United States, and the implications were derived through network analysis of keywords.