• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.391-395
• /
• 2021

텍스트 스타일 변환은 입력 스타일(source style)로 쓰여진 텍스트의 내용(content)을 유지하며 목적 스타일(target style)의 텍스트로 변환하는 문제이다. 텍스트 스타일 변환을 시퀀스 간 변환 문제(sequence-to-sequence)로 보고 기존 기계학습 모델을 이용해 해결할 수 있지만, 모델 학습에 필요한 각 스타일에 대응되는 병렬 말뭉치를 구하기 어려운 문제점이 있다. 따라서 최근에는 비병렬 말뭉치를 이용해 텍스트 스타일 변환을 수행하는 방법들이 연구되고 있다. 이 연구들은 주로 인코더-디코더 구조의 생성 모델을 사용하기 때문에 입력 문장이 가지고 있는 내용이 누락되거나 다른 내용의 문장이 생성될 수 있는 문제점이 있다. 본 논문에서는 마스크 언어 모델(masked language model)을 이용해 입력 텍스트의 내용을 유지하면서 원하는 스타일로 변경할 수 있는 텍스트 스타일 변환 방법을 제안하고 한국어 긍정-부정, 채팅체-문어체 변환에 적용한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07a
• /
• pp.688-690
• /
• 2005

본 논문은 2D 실사 기반 가상 헤어스타일의 자연스런 착용감을 제공하기 위한 것으로, 가상 헤어스타일러의 모발 블렌딩 방법에 관한 것이다. 제안된 방법을 2D 실사 영상에서 추출한 헤어스타일을 임의의 인물 영상의 두상에 정렬시킨 상타에서 원래의 헤어스타일에서 추출한 헤어스타일로 변해가는 반자동 필드 모핑을 수행할 시에 머리모양의 변형을 따라 단계적으로 변하는 가변 모핑 마스크를 사용하여 가상 헤어스타일을 생성하고 이렇게 생성된 헤어스타일을 가변 모핑 마스크의 경계 영역에서 시그모이드 함수에 기반한 모발 블렌딩을 수행함으로써 이중 노출 현상이 제거됨과 동시에 자연스러운 착용감을 제공하는 여러 유형의 헤어스타일을 자동으로 생성할 수 있다. 통상 이러한 작업은 사용자의 수작업에 전적으로 의존하는 반면에 제안된 방법은 반자동 필드 모핑에 기반한 편리한 사용자 인터페이스를 제공할 수 있기 때문에 반자동화를 통해 작업 피로도와 작업 시간을 경감시킬 수 있고 비숙련자도 간단한 사용자 입력을 통해 이중 노출 현상이 제거됨과 동시에 자연스런 착용감을 제공하는 가상 헤어스타일을 생성할 수 있는 이점이 있다.

• Journal of Convergence for Information Technology
• /
• v.11 no.9
• /
• pp.208-213
• /
• 2021

This study is a proposal for art mask design through the use of Art Nouveau. Based on precedent studies, its characteristics have been divided into a curvy expression, feminine expression, naturalism expression, and a repetitive expression. The four pieces by Gustav Klimt, a signature artist of Art Nouveau, have been selected. With the motif from them, the art masks were redesigned and the following results have been obtained. First, the study sought out a possibility and the methodologies to apply Art Nouveau to the art mask design. Second, the study has improved the accessibility to materials by increasing the scopes and methods of the art masks. Third, an integration of the art masks with the art history has provided a suitable condition for them. Based on such study results, it is expected that art mask materials could be developed and their status, as a field of makeup, could be raised.

• Proceedings of the Korea Contents Association Conference
• /
• 2008.05a
• /
• pp.3-9
• /
• 2008

This paper is related to hair-blending for 2D virtual color hairstyler based on the semi-automatic field morphing. The proposed 2D virtual color hair-styler is characterized as generating virtual hairstyles using the variable morphing mask instead of the fixed one for the semi-automatic morphing, and then hair-blending the generated virtual hairstyles using sigmoid function around the boundary region of the variable morphing mask. The proposed method is able to generate the virtual hairstyles with an easy-to-use interface based on semi-automatic field morphing. According to the proposed method, a user can shorten the working time and even an unskilled user can obtain natural hairstyles as he or she designates a small number of external user inputs.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.358-360
• /
• 2020

본 팀은 웹캠으로 촬영한 영상에서 원하는 물체를 선택하여 텍스처를 선택한 이미지의 스타일로 변환하는 프로젝트를 수행했다. 영상을 세그멘테이션하고 원하는 물체만을 원하는 텍스처로 변환하여 최종 아웃풋을 얻는다. 제안하는 네트워크는 물체를 다양한 스타일로 바꾸는 것이 가능한데, 이 중에서 이미지에 명화의 화풍을 입히는 것을 중점으로 하여 데모를 구현했다. 빠른 속도로 네트워크를 실행하기 위해 기존 연구들에 비디오 처리의 관점을 접목했다. 여러 프레임을 묶어 옵티컬 플로우를 생성하고, 첫 번째 프레임을 인스턴스 세그멘테이션한 후 마스크를 추출했다. 이후 마스크 영역만 뽑아낸 이미지를 새로운 입력으로 하여 스타일 트랜스퍼를 거치고, 이 첫번째 프레임과 나머지 프레임들의 옵티컬 플로우로 나머지 프레임들의 세그멘테이션과 스타일 트랜스퍼를 예측하여 다시 비디오 프레임으로 만들어 주었다. 본 알고리즘은 옵티컬 플로우 설정으로 네트워크의 계산량을 줄이며 속도를 개선했다. 빠른 데이터 처리로 사용자가 원하는 물체의 텍스쳐가 바뀔 수 있게 되었고, 이는 현실 세계가 실제로 바뀐 듯한 느낌을 들게 한다. 또한, 컴퓨터 비전에서 활발하게 연구되었던 분야를 AR로 끌어와 두 분야의 융합 가능성을 열었다. 현재 코로나의 영향으로 집에서 취미생활을 즐기는 인구가 많아졌다. 본 연구를 통해 많은 사람에게 집에서 쉽게 명화의 감성을 즐기고 느낄 수 있는 양질의 콘텐츠를 제공해주려 한다. 또한, 박물관과 미술관 등의 기관에서도 이 기술이 활용될 수 있다. 명화를 느낄 수 있는 다양한 콘텐츠를 이용하여 박물관이나 미술관의 홍보 효과도 기대할 수 있다.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.48 no.3
• /
• pp.417-432
• /
• 2024

This study aimed to enhance the comfort of health masks by conducting in-depth interviews, online surveys, and wearing tests on commercial health masks. The findings are summarized below. In-depth interviews revealed that the most comfortable mask styles were ranked as follows: Beak-1 (B-1) & Beak-2 (B-2) style > Flat style > 3-Horizontal foldable (3H) style. Men generally preferred flat style masks, while women favored beak-shaped masks. The internet survey results showed that 77.8% of surveyed brands offered a variety of mask styles, with 3H and B-2 masks being the most common. Different brands provided masks with different filtration levels, ranging from KF-AD to KF94. Size consistency also varied among brands, with flat masks having relatively consistent dimensions and B-2 masks showing significant size differences. Wearing tests indicated that 3H and B-1 masks were highly satisfying for categories like "itchiness," "unpleasant odor," and "mask slipping." Conversely, flat masks scored the highest satisfaction in classifications such as "stiff," "heavy," "thick," "hot while wearing it," "moisture accumulation," "breathing discomfort," "short ear straps," and "itchiness". Overall, the flat style, B-1, and B-2 yielded higher satisfaction levels, while 4-Horizontal foldable (4H) and 3H garnered lower satisfaction scores on wearing tests.

• Journal of Broadcast Engineering
• /
• v.27 no.5
• /
• pp.619-628
• /
• 2022

Although face recognition and face generation have been growing in popularity, the privacy issues of using facial images in the wild have been a concurrent topic. In this paper, we propose a face editing network that can reduce privacy issues by generating face images with various properties from a small number of real face images and facial mask information. Unlike the existing methods of learning face attributes using a lot of real face images, the proposed method generates new facial images using a facial segmentation mask and texture images from five parts as styles. The images are then trained with our network to learn the styles and locations of each reference image. Once the proposed framework is trained, we can generate various face images using only a small number of real face images and segmentation information. In our extensive experiments, we show that the proposed method can not only generate new faces, but also localize facial attribute editing, despite using very few real face images.

• Journal of Korea Entertainment Industry Association
• /
• v.14 no.5
• /
• pp.289-297
• /
• 2020

The purpose of this study is to learn about the life style change experience of undergraduate due to COVID-19. The study was conducted from March 23, 2020 to March 30, 2020 and was conducted on five undergraduate residing in Jeonju. We valued individual's subjective experiences and analyzed them by applying on of the qualitative studies to identify and describe the phenomena as they are. The study found that the lifestyle changes of undergraduate due to COVID-19 were derived from three components: Psychological change, Environmental change, and Behavioral change. Though the participants felt uneasy and terrified by COVID-19 through the experience of psychological change, they expressed relief and anticipation for the increase in the number of complete healers. Experience of environmental change has experienced social distance, changes in academic operation, and difficulty in finding jobs to prevent infection. Behavioral changes have experienced a matrix of mask purchases and thorough management of personal hygiene, which is considered to protect oneself from COVID-19 while also being considerate to others. It is believed that the government will need to provide psychological and behavioral mediation programs for problems arising from the practice of social distance.

• The Journal of the Korea Contents Association
• /
• v.9 no.12
• /
• pp.86-93
• /
• 2009

In this paper, we propose a method that generates a sketch likely to be drawn by a human from a real image(a computer generated image or a photo image). The method can create human friendly images with pen strokes by using image processing techniques such as spatial filters. In order to create the human friendly image, first, we made the tiny unnecessary contours. Second, we made the second image by inverting the first image and performed a processing similar to the first processing. Last, we subtracted the second image from the first image and added a hatch to it. The sequence of processing enables us to generate a human friendly image with line shading, likely to be sketched by water pen compared to the past method(pencil hatching and sketch effects). Also the proposed method can represent the sketch effects more effectively with applying different hatch to the shading area of a input image.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.3
• /
• pp.1-8
• /
• 2022

Recently, Google MedeiaPipe presents a novel approach for neural network-based hair segmentation from a single camera input specifically designed for real-time, mobile application. Though neural network related to hair segmentation is relatively small size, it produces a high-quality hair segmentation mask that is well suited for AR effects such as a realistic hair recoloring. However, it has undesirable segmentation effects according to hair styles or in case of containing noises and holes. In this study, the energy function of the test image is constructed according to the estimated prior distributions of hair location and hair color likelihood function. It is further optimized according to graph cuts algorithm and initial hair region is obtained. Finally, clustering algorithm and image post-processing techniques are applied to the initial hair region so that the final hair region can be segmented precisely. The proposed method is applied to MediaPipe hair segmentation pipeline.