• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.643-646
• /
• 2020

본 논문에서는 직전 연구였던 비트 평면 분할과 디더링을 통한 다중 학습 통합 신경망 구축에서의 한계점을 분석하고, 향상시킨 방법을 제시한다. 통합 신경망을 구축하는 방법에 대해 최근까지 시도되었던 방법들은 신경망을 구성하는 가중치(weight)나 층(layer)를 공유하거나 태스크 별로 구분하는 것들이 있다. 이와 같은 선상에서 본 연구는 더 작은 단위인 가중치의 비트 평면을 태스크 별로 할당하여 보다 효율적인 통합 신경망을 구축한다. 실험은 이미지 분류 문제에 대해 수행하였다. 대중적인 신경망 구조인 ResNet18 에 대해 적용한 결과 데이터셋 CIFAR10 과 CIFAR100 에서 이론적인 압축률 50%를 달성하면서 성능 저하가 거의 발견되지 않았다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.83-84
• /
• 2023

2D 게임의 이미지들은 2D 스프라이트(Sprite) 조각들을 같은 평면에 겹쳐 그리는 방식으로 표현한다. 탑다운 뷰(Top-Down View) 2D 게임 시점은 평면의 그림에 입체적인 묘사를 함으로써 캐릭터나 오브젝트가 수직으로 일어서 있는듯한 3D 느낌을 전달한다. 그러나 실제로는 2D 평면이므로 3D 그림자 맵(Shadow Map) 방식을 사용할 수 없는 단점이 있다. 본 논문에서는 2D 스프라이트 오브젝트의 그림자를 3D 그림자맵으로 생성하는 방법과 동반되는 이슈들을 제시한다.

• Smart Media Journal
• /
• v.5 no.1
• /
• pp.61-68
• /
• 2016

According to increasing SNS users and developing smart devices like smart phone and tablet PC recently, many techniques to classify user emotions with social network information are researching briskly. The use emotion classification stands for distinguishing its emotion with text and images listed on his/her SNS. This paper suggests a method to classify user emotions through sampling a value of a representative figure on a trigonometrical function, a representative adjective on text, and a canny algorithm on images. The sampling representative adjective on text is selected as one of high frequency in the samplings and measured values of positive-negative by SentiWordNet. Figures sampled on images are selected as the representative in figures; triangle, quadrangle, and circle as well as classified user emotions by measuring pleasure-unpleased values as a type of figures and inclines. Finally, this is re-defined as x-y graph that represents pleasure-unpleased and positive-negative values with wheel of emotions by Plutchik. Also, we are anticipating for applying user-customized service through classifying user emotions on wheel of emotions by Plutchik that is redefined the representative adjectives and figures.

• Proceedings of the Optical Society of Korea Conference
• /
• 2006.02a
• /
• pp.119-120
• /
• 2006

• Journal of KIISE
• /
• v.42 no.2
• /
• pp.235-241
• /
• 2015

Locally linear embedding (LLE) [1] is a type of manifold algorithms, which preserves inner product value between high-dimensional data when embedding the high-dimensional data to low-dimensional space. LLE closely embeds data points on the same subspace in low-dimensional space, because the data points have significant inner product values. On the other hand, if the data points are located orthogonal to each other, these are separately embedded in low-dimensional space, even though they are in close proximity to each other in high-dimensional space. Meanwhile, it is well known that the facial images of the same person under varying illumination lie in a low-dimensional linear subspace [2]. In this study, we suggest an improved LLE method for face recognition problem. The method maximizes the characteristic of LLE, which embeds the data points totally separately when they are located orthogonal to each other. To accomplish this, all of the subspaces made by each class are forced to locate orthogonally. To make all of the subspaces orthogonal, the simultaneous Diagonalization (SD) technique was applied. From experimental results, the suggested method is shown to dramatically improve the embedding results and classification performance.

• Journal of the Korean Geotechnical Society
• /
• v.32 no.3
• /
• pp.49-60
• /
• 2016

Soil failure is initiated and preceded by forming and progressing of shear band, defined as the localization of deformation into thin zones of soil mass. To understand the failure mechanism of normally consolidated cohesive soil, the spatial distribution and evolution of deformation within the entire specimen need to be evaluated. In this study, vertical compression tests under plane strain condition were performed on reconstituted kaolinite specimens, while capturing digital images of the specimen at regular intervals during shearing. Overall stress-strain behavior from initial to post peak has been analyzed together with spatial distributions of deformations and shear band characteristics from digital images at 4 stages.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.44-46
• /
• 2022

Various object recognition using artificial intelligence basically shows planar results. It is based on classifying objects or identifying what objects are on the image. However, the original object has a three-dimensional shape, not a plane, and although the perception to obtain only simple results from the image does not matter, there is a lot of information that is insufficient when used in various fields. In this paper, checks the method of generating data in various fields of objects and whether it is meaningful by utilizing the characteristics of Layer that generates intermediate results with respect to image generation based on the GAN algorithm. It solves some of the problems in the hardware and collection process for generating existing multi-faceted data, and confirms that it can be utilized after data generation on several limited objects.

• The Journal of the Convergence on Culture Technology
• /
• v.5 no.3
• /
• pp.283-288
• /
• 2019

We propose a method to search for similar images by using outline lines and viewpoints. In the first test, the three-dimensional image, which can't control the motion, has lower search accuracy than the static flat image. For the cause analysis, six specific tropical fish data were selected. We made a 3D graphics of tropical fishes of each kind, and we made 144 image outline lines with 12 stage viewpoints of top, bottom, left and right. Tropical fish by type were collected and sorted by time of search through similar search. Studies have shown that there are many unique viewpoints for each species of tropical fish. To increase the accuracy of the search, a User Interface was created to select the user's point of view. When the user selects the viewpoint of the image, a method of showing the result in consideration of the range of the related viewpoint is proposed.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2011.07a
• /
• pp.589-591
• /
• 2011

채색화 기법은 일부 픽셀의 색상 정보를 이용하여 흑백의 이미지에 색상 정보를 추가하는 기법이다. 이러한 채색화 기법을 기반으로한 색상 이미지 압축기법들이 연구되고 있다. 색상 평면에서 대표적인 픽셀들을 소스 픽셀로 자동적으로 선택하고, 이 소스 픽셀들의 위치와 색상 정보만을 디코더에 압축하여 전송한다. 본 논문에서는 밝기 변화량을 이용하여 소스 픽셀의 위치를 결정함으로써, 디코더에서도 동일한 작업으로 소스 픽셀의 위치를 결정할 수 있다. 따라서 소스 픽셀에 대한 위치정보를 전송하기 위한 비트량을 줄임으로써 압축 효율을 높였다. 제안알고리듬은 디코더에서 색상정보의 복원에 이용하는 채색화 기법의 특성에 맞추어서 밝기가 평평하고 넓은 영역에서 먼저 소스픽셀을 선택하여, 이웃의 비슷한 밝기를 가지는 픽셀에 대한 색상 정보를 효율적으로 압축한다.

• Proceedings of the Korea Contents Association Conference
• /
• 2011.05a
• /
• pp.155-156
• /
• 2011

기존의 디스플레이는 2차원적인 평면 이미지만을 제공하고 있으며, 3차원 이미지를 제공하기 위해서는 추가적인 장비를 설치해야 한다. 본 논문에서는 일반적으로 널리 쓰이는 LCD 디스플레이 내부에 3차원 입체적인 이미지를 제공하는 방법을 제시하고자 한다. 이를 위해 사용자의 위치를 추적할 때 적외선 카메라를 사용함으로써 사용자는 별도의 장비를 사용하지 않아도 된다. u-City 환경에서 이와 같은 방법으로 GIS 정보가 LCD 디스플레이 내부에 입체적인 형태로 보다 효과적으로 보여진다.