• 복식문화연구
• /
• 제9권3호
• /
• pp.412-432
• /
• 2001

Internet search engines are used by the majority of find information on the Web. However, Web users can be often dissatisfied with the mistakes in the retrieval of ‘Fashion’ information from the Internet. The purpose of this study is to analyze the ‘Fashion’ category structure in the Internet search engines. There are 2 steps for achieving it: the first, to investigate the structures of ‘Fashion’ categories and then, to analyze the gap between ‘Fashion’ categories defined by them and extensive ‘Fashion’categories, which are approached on 2 sides of the fashion-life and fashion-business. We select 5 major search engines for the case study: Yahoo, Lycos, Naver, Hanmir, Empas, which ranked as top 5 of total search engines and potal sites in February, 2001, and retrieve ‘Fashion’ categories from the first level to the last level by using both “topics retrieval”. Eventually, we can find the problems of ‘Fashion’ category structure in search engines. Also, it is concluded with a brief perspective of ‘Fashion’ categories in the Internet search engines and the implications for the future.

• 디지털융복합연구
• /
• 제16권12호
• /
• pp.301-307
• /
• 2018

검색에서 이미지는 시각적 속성이 중요지만, 기존의 검색방법은 문서 검색을 위한 방법에 초점이 맞춰져 있어 이미지의 속성 정보가 미반영된 키워드 중심의 검색 시스템이 대부분이다. 본 연구는 이러한 한계를 극복하고자 이미지의 벡터정보를 기반으로 유사 이미지를 검색할 수 있는 모델과 스케치로 검색 쿼리를 제공하여 유사 이미지를 검색할 수 있는 시스템을 개발하였다. 제안된 시스템은 GAN을 이용하여 스케치를 이미지 수준으로 업 샘플링하고, 이미지를 CNN을 통해 벡터로 변환한 후, 벡터 공간 모델을 이용하여 유사 이미지를 검색한다. 제안된 모델을 구현하기 위하여 패션 이미지를 이용하여 모델을 학습시켰고 패션 이미지 검색 시스템을 개발하였다. 성능 측정은 Precision at k를 이용하였으며, 0.774와 0.445의 성능 결과를 보였다. 제안된 방법을 이용하면 이미지 검색 의도를 키워드로 표현하는데 어려움을 느끼는 사용자들의 검색 결과에 긍정적 효과가 나타날 것으로 기대된다.

• 지능정보연구
• /
• 제28권4호
• /
• pp.329-346
• /
• 2022

최근 소셜 미디어의 숏폼(Short form) 동영상(인스타그램, 틱톡, 유튜브) 시장이 점차 증가하면서 인공지능 영역에서는 이를 활용한 연구가 활발히 진행되고 있다. 대표적인 연구분야로 동영상 내의 패션 상품을 탐지하고 상품 이미지를 검색하는 Video to shop 을 들 수 있다. 이와 같은 동영상 기반 인공지능 모델에서는 Convolution 연산을 사용하여 상품의 특징을 추출한다. 하지만 연산 자원의 제한으로 인해, 동영상의 모든 프레임을 사용하여 특징을 추출하는 것은 현실적으로 불가능하다. 이로 인해, 기존 연구에서는 전체 프레임 중 일부만 샘플링해서 사용하거나, 주제의 특성을 활용한 샘플링 방법을 개발하여 이를 통해 위 문제점을 개선하고, 모델의 성능도 향상시켰다. 기존의 Video to shop 연구에서는 프레임을 샘플링 할 때, 무작위로 일부분의 프레임을 샘플링하거나 균등한 간격으로 샘플링 한다. 하지만 이러한 샘플링 방법은 상품이 존재하지 않는 노이즈 프레임을 샘플링 하면서 패션 상품 검색 모델의 성능을 저하시킨다. 이에 본 연구는 노이즈 프레임을 제거하고 검색 모델의 성능을 향상시키는 샘플링 방법 MF(Missing Fashion items on frame) sampler를 제안한다. MF sampler는 키 프레임 메커니즘(Mechanism)을 발전시켜 자원 한계의 문제점을 개선했다. 또한, 노이즈 탐지 모델을 활용한 노이즈 프레임 제거를 통해 검색 모델의 성능을 향상시켰다. 이와 같은 결과는 실험을 통해 확인되었고, Video to shop 패션 상품 검색에 있어 성능 향상과 효과적인 학습이 가능하다는 것을 확인할 수 있었다.

• 인터넷정보학회논문지
• /
• 제21권1호
• /
• pp.33-43
• /
• 2020

Retrieving a 3D model from a 3D database and augmenting the retrieved model in the Augmented Reality system simultaneously became an issue in developing the plausible AR environments in a convenient fashion. It is considered that the sketch-based 3D object retrieval is an intuitive way for searching 3D objects based on human-drawn sketches as query. In this paper, we propose a novel deep learning based approach of retrieving a sketch-based 3D object as for an Augmented Reality Model. For this work, we introduce a new method which uses Sketch CNN, Wasserstein CNN and Wasserstein center loss for retrieving a sketch-based 3D object. Especially, Wasserstein center loss is used for learning the center of each object category and reducing the Wasserstein distance between center and features of the same category. The proposed 3D object retrieval and augmentation consist of three major steps as follows. Firstly, Wasserstein CNN extracts 2D images taken from various directions of 3D object using CNN, and extracts features of 3D data by computing the Wasserstein barycenters of features of each image. Secondly, the features of the sketch are extracted using a separate Sketch CNN. Finally, we adopt sketch-based object matching method to localize the natural marker of the images to register a 3D virtual object in AR system. Using the detected marker, the retrieved 3D virtual object is augmented in AR system automatically. By the experiments, we prove that the proposed method is efficiency for retrieving and augmenting objects.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2003년도 봄 학술발표논문집 Vol.30 No.1 (C)
• /
• pp.253-255
• /
• 2003

Electricity is the prime commodity in mobile device, e.g. smart phone, PDA, MP3 player and etc. This strict restriction on power consumption requirement of the mobile device puts unique demand in designing hardware and software components of the device. In this paper, we address the issue of minimizing the power consumption in retrieving the continuous media data from the disk drive for real-time playback purpose. Different from the legacy text based data, real-time multimedia playback requires that the storage supplies the data block continuous fashion. This may put immense burden on the power scarce environment since the disk Is required to be active for the entire playback duration. We develop elaborate algorithm which carefully analyzes the power consumption profile of the disk drive and which establishes the data retrieval schedule for the given playback. It computes the amount of data blocks to read, the length of active and standby period. According to our simulation result, the ARM algorithm exhibits superior performance in continuous media retrieval from the aspect of power consumption to legacy playback scheme.

• 한국의상디자인학회지
• /
• 제20권4호
• /
• pp.57-71
• /
• 2018

Since consumers have become more comfortable with providing and receiving information online, 'online word of mouth' has been gaining consideration as one of the major information sources. Also, the shopping orientation of consumers has been proven to be an important determinant of consumer behavior. Therefore, the study investigated the differences in online WOM behavior based on shopping orientation. Hedonic, loyal, and syntonic styles were the types of shopping orientation considered, and the study focused on information retrieval tendencies, the motivation of online WOM search, searching online WOM sources, and the contents for the online WOM behavior. The research conducted an off-line survey targeting females in their twenties. The total number of data sets used in the empirical study was 125, and these were analyzed by SPSS 20.0: factors analysis, Cronbach's ${\alpha}$, k-means cluster, ANOVA, Duncan's multiple range test, Kruskal-Wallis, Mann-Whitney, and Bonferroni correction. The participants were divided into 3 kinds of shopping orientation groups named 'trend-pursuit', 'passive', and 'loyal'. As a result, there were significant differences in online WOM behavior discovered between the groups. Firstly, the 'trend-pursuit' group had the highest number of ongoing searches while the 'loyal' group had the highest number of pre-purchase search. Secondly, the 'trend-pursuit' and 'loyal' groups both had the motivations of online WOM search, hedonic and utility, whereas the 'passive' group had the lowest motivations for both motivations. Thirdly, the 'loyal' group frequently referred to reviews on shopping malls as online WOM sources. The research provided a better understanding of the online WOM behavior of present consumers and suggests that fashion related corporations map out marketing strategies with the understanding of these behaviors.

• 한국산업정보학회논문지
• /
• 제27권6호
• /
• pp.13-23
• /
• 2022

본 논문에서는 패션 분야의 비정형 데이터 검색을 위한 패션 아이템별 세부 컨포넌트 이미지 분류 알고리즘을 제안한다. 코로나-19 환경으로 인하여 최근 AI 기반 쇼핑몰이 증가하는 추세이다. 하지만 기존의 키워드 검색과 사용자 서핑 행위 기반 개인 맞춤형 스타일 추천으로는 정확한 비정형 데이터 검색에는 한계가 있다. 본 연구는 다양한 온라인 쇼핑 사이트에서 크롤링한 이미지를 사용하여 Mask R-CNN을 활용한 전처리를 진행한 후, CNN을 통해 패션 아이템별 컴포넌트에 대한 분류를 진행하였다. 셔츠의 카라 및 패턴과 청바지의 핏, 워싱 및 컬러에 대한 분류를 진행하였으며, 다양한 전이학습 모델을 비교 분석한 후 가장 높은 정확도가 나온 Densenet121모델을 사용하여 셔츠의 카라는 93.28%, 셔츠의 패턴은 98.10%의 정확도를 도달하였으며, 청바지의 핏은 Notched, Spread, Straight 3가지의 클래스의 경우 91.73%, Regular 핏을 추가한 4가지의 클래스의 경우 81.59%, 청바지의 색상은 93.91%, 청바지의 Washing은 91.20%, 청바지의 Demgae는 92.96%의 정확도를 도출하였다.

• 한국산학기술학회:학술대회논문집
• /
• 한국산학기술학회 2003년도 Proceeding
• /
• pp.57-64
• /
• 2003

Achieving the beneficiary goal of recent discovery in human genome project still needs a way to retrieve and analyze the exponentially expanding bio-related information. Research on bio-related fields naturally applies knowledge discovered to the current problem and make inferences to extract new information where shared concepts and data containing information need to be defined and used in a coherent way. In such a professional domain, while the need to help users reduce their work and to improve search results has been emerged. methods for systematic retrieval and adequate exchange of relevant information are still in their infancy. The design of our system aims at improving the quality of information retrieval in a professional domain by utilizing both corpus-based and concept-based ontology. Meta-rules of helping users to make an adequate query are formed into an ontology in the domain. The integration of those knowledge permits the system to retrieve relevant information in a more semantic and systematic fashion. This work mainly describes the query models with details of GUI and a secondary query generation of the system.

• 복식문화연구
• /
• 제24권2호
• /
• pp.248-262
• /
• 2016

Most Korean companies in the fashion industry are SMEs, and the role of the CEO and management ownership is important for enhancing the firm's competence and developing strategies. The study aims to examine the effect of management ownership on company financial growth. In particular, the study focuses on the moderating effect of company age and size on Korean fashion SMEs' financial outcomes. Financial data based on company financial statements from 2012 to 2014 was collected by the Data Analysis, Retrieval and Transfer System of Korea's Financial Supervisory Service. A total of 295 companies' (domestic fashion businesses) data was analyzed by the bootstrap method. The median sales value in the financial year 2014 was 47,492,403,958 KRW, and the company size was divided by it. The companies were in business for an average of 20 years. According to the results, the management ownership had a negative effect on Compound Annual Growth Rate (CAGR) for the three-years, and the relationship between the two variables was moderated by company age. Additionally, the interaction effect of management ownership and company age on 3-CAGR was also moderated by company size. When the companies had spent only a few years in business, a negative effect of management ownership for small firms and a positive effect of management ownership on financial growth for medium firms were found. These results suggest that small companies starting business need to manage their company governance structure to make flexible decisions, and after retaining financial growth, the companies can expand their businesses based on strong ownership.

• Journal of Information Processing Systems
• /
• 제14권2호
• /
• pp.469-486
• /
• 2018

We propose an enhanced version of the local binary pattern (LBP) operator for texture extraction in images in the context of image retrieval. The novelty of our proposal is based on the observation that the LBP exploits only the lowest kind of local information through the global histogram. However, such global Histograms reflect only the statistical distribution of the various LBP codes in the image. The block based LBP, which uses local histograms of the LBP, was one of few tentative to catch higher level textural information. We believe that important local and useful information in between the two levels is just ignored by the two schemas. The newly developed method: gradual locality integration of binary patterns (GLIBP) is a novel attempt to catch as much local information as possible, in a gradual fashion. Indeed, GLIBP aggregates the texture features present in grayscale images extracted by LBP through a complex structure. The used framework is comprised of a multitude of ellipse-shaped regions that are arranged in circular-concentric forms of increasing size. The framework of ellipses is in fact derived from a simple parameterized generator. In addition, the elliptic forms allow targeting texture directionality, which is a very useful property in texture characterization. In addition, the general framework of ellipses allows for taking into account the spatial information (specifically rotation). The effectiveness of GLIBP was investigated on the Corel-1K (Wang) dataset. It was also compared to published works including the very effective DLEP. Results show significant higher or comparable performance of GLIBP with regard to the other methods, which qualifies it as a good tool for scene images retrieval.