• Journal of the Korean Society of Clothing and Textiles
• /
• v.31 no.12
• /
• pp.1784-1792
• /
• 2007

This study set out to identify the problems with hat labels and to search for improvement measures by examining and analyzing consumers' practice of managing their hats. It also intended to provide accurate and enough information about how to keep and wash hats and thus help consumers use their hats for a long period. In an attempt to investigate how consumers wash and manage their hats, a survey was carried out to 395 individuals in their twenties and over who owned hats living in urban areas including Seoul, and were quota sampled according to age and gender. The survey period is March to April 2007. The collected data were statistically treated with the SPSS 12.0 program in terms of frequency, percentage, mean, standard error, cross tabulation, t-test, and one-way ANOVA. The findings were as followed. First, the respondents were in the average level of perceiving and practicing the washing methods of their hats. The female respondents who had more experiences with laundering than the males knew and practiced the washing methods for hats better than males. Second, compared to other clothing items, hat wearers were more likely to pay careful attention to their hats by putting their hats in a laundry net and applying a laundry detergent for wool fabrics when using a washing machine or washing their hats with their own hands. And third, most of the hat wearers were aware of the importance of hat labels and showed a lower level of trust in them than other clothing items. The suppliers need to offer accurate and practical labels in order to regain the consumers' trust. Many consumers had some difficulties figuring out the size system of hats. In particular, the male consumers had a low level of perception of labels, which implies that there should be specific efforts to educate them about general labels.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.7
• /
• pp.816-822
• /
• 2016

Content-Based image retrieval is a method to search by image features such as local color, texture, and other image content information, which is different from conventional tag or labeled text-based searching. In real life data, the number of images having tags or labels is relatively small, so it is hard to search the relevant images with text-based approach. Existing image search method only based on image feature similarity has limited performance and does not ensure that the results are what the user expected. In this study, we propose and validate a machine learning based approach to improve the performance of the image search engine. We note that when users search relevant images with a query image, they would expect the retrieved images belong to the same category as that of the query. Image classification method is combined with the traditional image feature similarity method. The proposed method is extensively validated on a public PASCAL VOC dataset consisting of 11,530 images from 20 categories.

• Proceedings of the KAIS Fall Conference
• /
• 2007.05a
• /
• pp.181-184
• /
• 2007

자연영상에 포함된 텍스트는 많은 중요한 정보를 포함하고 있으므로 자연영상에서 텍스트 정보를 검출하는 연구가 활발히 진행되고 있다. 본 논문에서는 문자 영역의 구조적인 특정을 배열문법으로 정의한 적응적 문자-에지 맵을 제안하여 텍스트 영역을 검출한다. 캐니-에지 검출기로 에지를 추출하고, 생성된 에지 이미지를 레이블링하고 그 영역의 문자구조 특징을 분석하기 위해서 적응적 문자-에지 맵을 분석한다. 적응적 문자-에지 랩의 분포 상태를 분석함으로서 텍스트 후보 영역을 검출하고, 텍스트 영역의 에지 히스토그램 프로파일을 분석함으로서 텍스트 후보 영역에 대한 검증을 수행하여 최종적인 텍스트 영역을 검출한다. 제안한 방법은 다양한 종류의 자연영상을 대상으로 실험하였고, 기울어진 텍스트와 다양한 크기의 텍스트 구성된 자연영상에서 텍스트 영역을 효과적으로 검출하였다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1995.06a
• /
• pp.265-271
• /
• 1995

한국전자통신연구소의 음성 데이터베이스 구축 현황을 기술한다. 현재 한국전자통신연구소에서는 음성인식 연구를 위해 단어음성, 정형 문장 음성 등의 데이터베이스를 구축, 보유하고 있다. 음성인식용 데이터베이스는 정해진 단어, 분장을 20명 내지 100명이 발성한 것으로, 일부는 음소 단위까지 레이블링이 되어 있다. 또 음성합성 연구를 위해 합성단위 및 운율데이타베이스를 가지고 있는데 이는 한 명 혹은 남녀 각각 3명이 발성한 것으로, 일부는 피치 등이 수록되어 있다. 문장 데이터베이스는 언어 정보처리를 위해 교재, 문학, 경제, 과학 분야의 문장을 총 480만 어절 가지고 있으며, 이 중 일부에 품사 정보를 추가하였다. 한국전자통신연구소는 국내 음성 연구의 발전에 기여하고자 음성 연구의 기반 자료가 되는 음성 데이터베이스를 국내 대학 및 산업체에 배포하고 있다 음성 데이터베이스는 음성 연구의 기반이 되는 자료임에도 불구하고 많은 비용과 노력이 들어 일반 대학에서는 쉽게 만들 수 없었다. 이에 ETRI는 한국통신이 출연한 "자동통역 요소기술개발" 과제으 LQNTKSANF인 여러 종류의 음성 데이터베이스와 관련 프로그램을 공급하여 국내 음성 연구의 기반 확립에 기여하고자 한다. 기여하고자 한다.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.479-487
• /
• 2021

In this paper, we propose an Sound Event Detection (SED) model using self-training based on a noisy student model. The proposed SED model consists of two stages. In the first stage, a mean-teacher model based on an Residual Convolutional Recurrent Neural Network (RCRNN) is constructed to provide target labels regarding weakly labeled or unlabeled data. In the second stage, a self-training-based noisy student model is constructed by applying different noise types. That is, feature noises, such as time-frequency shift, mixup, SpecAugment, and dropout-based model noise are used here. In addition, a semi-supervised loss function is applied to train the noisy student model, which acts as label noise injection. The performance of the proposed SED model is evaluated on the validation set of the Detection and Classification of Acoustic Scenes and Events (DCASE) 2020 Challenge Task 4. The experiments show that the single model and ensemble model of the proposed SED based on the noisy student model improve F1-score by 4.6 % and 3.4 % compared to the top-ranked model in DCASE 2020 challenge Task 4, respectively.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.10
• /
• pp.1010-1014
• /
• 2010

Article-related product recommender system is an emerging e-commerce service which recommends items based on association in contexts between items and articles. Current services recommend based on the similarity between tags of articles and items, which is deficient not only due to the high cost in manual tagging but also low accuracies in recommendation. As a component of novel article-related item recommender system, we propose a new method for tagging item images based on pre-defined categories. We suggest a hypernetwork-based algorithm for learning association between images, which is represented by visual words, and categories of products. Learned hypernetwork are used to assign multiple tags to unlabeled item images. We show the ability of our method with a product set of real-world online shopping-mall including 1,251 product images with 10 categories. Experimental results not only show that the proposed method has competitive tagging performance compared with other classifiers but also present that the proposed multi-tagging method based on hypernetworks improves the accuracy of tagging.

• Food Science of Animal Resources
• /
• v.27 no.4
• /
• pp.392-400
• /
• 2007

This experiment was conducted to evaluate effects of various periods of rye silage feeding on the growth performance, blood characteristics, and carcass quality of finishing pigs. A total of sixteen [($Landrace{\times}Yorkshire{\times}Duroc$)] pigs (90.26 kg in average initial body weight) were tested in individual cages for a 30 day period. Dietary treatments included 1) CON (basal diet), 2) S10 (basal diet for 20 days and 3% rye silage for 10 days) 3) S20 (basal diet for 10 days and 3% rye silage for 20 days) and 4) S30 (3% rye silage for 30 days). There were no significant differences in the ADG and gain/feed ratio among the treatments(p>0.05), however the ADFI was higher in pigs fed the CON diet than with pigs fed diets with rye silage (p<0.05). The DM digestibility was higher with the S20 diet than with the S30 diet (p<0.05). With regard to blood characteristics, pigs fed rye silage had a significantly reduced cortisol concentration compared to pigs fed the CON diet (p<0.05). The backfat thickness was higher with the CON diet than with the S20 or S30 diets (p<0.05). Regarding the fatty acid contents of the leans, the C18:0 and total SFA were significantly higher with the CON diet than with the other diets (p<0.05). However, the C18:1n9, total MUFA and UFA/SFA levels were significantly lower with the CON diet than the other diets (p<0.05). Regarding the fatty acid contents of fat, the levels of C18:1n9 and MUFA were similar with the S20 and S30 diets, however, these levels were higher than with the CON or S10 diets (p<0.05). In conclusion, feed intake and DM digestibility were affected by rye silage, and the cortisol concentration, backfat thickness and fatty acid composition of pork were positively affected by feeding pigs rye silage.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.139-147
• /
• 2021

This paper introduces and implements a Sound Event Detection (SED) system based on weakly-supervised learning where only part of the data is labeled, and analyzes the effect of parameters. The SED system estimates the classes and onset/offset times of events in the acoustic signal. In order to train the model, all information on the event class and onset/offset times must be provided. Unfortunately, the onset/offset times are hard to be labeled exactly. Therefore, in the weakly-supervised task, the SED model is trained by "strongly labeled data" including the event class and activations, "weakly labeled data" including the event class, and "unlabeled data" without any label. Recently, the SED systems using the mean-teacher model are widely used for the task with several parameters. These parameters should be chosen carefully because they may affect the performance. In this paper, performance analysis was performed on parameters, such as the feature, moving average parameter, weight of the consistency cost function, ramp-up length, and maximum learning rate, using the data of DCASE 2020 Task 4. Effects and the optimal values of the parameters were discussed.

• Proceedings of the KAIS Fall Conference
• /
• 2007.11a
• /
• pp.262-265
• /
• 2007

이미지 데이터베이스 시스템에서 이미지에 포함된 문자정보를 기반으로 검색어를 사용한다면 검색의 정확도 높일 수 있다. 이미지에서 문자정보를 추출을 위한 전단계로서 문자열 영역 검출이 필수적인 과제가 된다. 그러므로 본 논문에서는 문자의 크기와 문자열의 방향에 적응적인 문자-에지 맵을 이용한 문자열 영역 검출 방법을 제안한다. 캐니-에지 검출기로 에지를 추출하고, 생성된 에지 이미지로 레이블 이미지를 얻고, 그 영역의 문자구조 특징을 분석하기 위해서 배열문법으로 문자-에지 맵에 적응적으로 분석한다. 문자-에지 맵의 분석결과로서 문자열 후보 영역을 얻고, 문자열 영역의 구조적인 특징을 이용하여 문자열 후보 영역을 검증함으로서 최종적인 문자열 영역을 검출한다. 제안한 방법은 다양한 종류의 자연영상을 대상으로 실험하였고, 자연영상에서 기울어진 문자열과 다양한 크기의 문자를 갖는 문자열 영역을 효과적으로 검출하였다.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.1
• /
• pp.251-256
• /
• 2021

This study analyzed the most frequently used navigation cases in a desktop environment. As a result of the research, GNB induces users' search as the top element of the search structure and can place color, text, icon, and image elements. LNB could be classified in the form of a dropdown, flyout, dropline and mega menu. In this study, the navigation structure of Interpark and Interpark among open markets used by users was analyzed. G-Market's GNB has a two-tier structure with color, text, image, and icon elements, and Interpark has a three-tiered horizontal label. Interpark's GNB drew attention by placing a badge on the seasonal label, which is a temporary content section, unlike G-market. It can be seen that the LNBs of both shopping malls have flyout menus that protrude when you mouse over the category menu arranged in a vertical text form under the logo placed on the left. The flyout menu has a complex structure consisting of the layout of the mega menu. This study is meaningful in revealing user experience elements by analyzing the GNB and LNB of shopping malls these days where internet shopping is increasing.