• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.

• Fashion & Textile Research Journal
• /
• v.14 no.5
• /
• pp.766-777
• /
• 2012

The 13 genre paintings by Junkeun Kim in the book "Korean Games(by Stewart Culin 1858~1929)" were used to study the late $19^{th}$ century's basic costumes and Games style in Korea. The people who appear in the paintings are 26 adult males, 2 minor males, 2 adult females and 3 kisaengs. Typical men wore 'Jeogori' which had various colors and white linings with a reached hip line, and knotted with a 'go-rum' on the right side. They also wore, white colored 'Baji' with colorful sash that were knot below knee or ankle together with 'Hang-jun' or 'Daenim'. They wore 'Beoseon'. Some men wore 'Po' whose colors were blue, green, indigo, white. The general women wore 'Jeogori' in deep green and light pink, indigo, green, red, and they matched with colors for 'Kit' and, 'Go-rum', 'Kut-dong' and its 'Go-rum' was short and narrow. It was so fit and short with narrow sleeve. It had 'Dunggun-kit'(round head collar) with white 'Dong-jung' and so it fit at neck. They wore 'Chi-ma' whose color was red, light green, or light indigo. It contrasted with 'Jeogori'. The width of 'Chi-ma' was big enough. Then white inner slacks came out under the skirt. Traditional Korean games can be classified according to age and gender. Then the games can be further classified into three categories : men's games, women's games, and games for all. The games for adult are an archery practice, hunting, shovel work with a karae, making a bow, drawing Jongkung-chart, Korean chess, playing paduk, and the Korean card game. A swing is a game for women. Games for both men and women are dice play, and domino game. Games for both adult and minor males are sledge, and tightrope walking. Through genre paintings in the $19^{th}$ century, I reached a conclusion that basic costumes are similar to 'Hanbok' at the present time and the method of wearing them has not changed much. It appears that the originality of traditional costumes has been maintained.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.32 no.9
• /
• pp.1333-1342
• /
• 2008

This study intended to evaluate the visual sensibility of textiles of knitting simulation through the painting function of SDS-ONE, a computer knitting system, with the use of various fancy yarns for hand knitting. For this study, six kinds of fancy yarns with difference manufacturing style and form were selected: loop yarn, tamtam yarn, fur1 yarn, fur2 yarn, ladder yarn, and tape yarn. In order to create a computer knitting simulation, paint function of SDS-ONE was employed to scan the fancy yarns for hand knitting with a scanner. Then, after the registration of the yarns as data, the yarn was brought in for knitting simulation in plain stitch with 3 gauge. After that a survey was conducted involving 60 female experts in their twenties. Using SPSS 12.0, factor analysis, reliability, Pearson correlation coefficient, ANOVA, Duncan test, and regression analysis were performed to analyze the collected data. First, according to the result of analyzing the visual sense factor, three factors of the 'inflexibleness/ruggedness sense', 'lightweight sense', and 'smoothness sense' was extracted. According to the result of analyzing the visual sensibility factor, three factors of the 'activity', 'grace', and 'purity' was extracted. Second, With regard to preference and purchase intention, the knitted fabric with tamtam yarn(H2) was preferred than the other samples, indicating that the respondents had intention to purchase it. As for word-of-mouth intention, H2 ranked the first place, implying the respondents had intention to recommend it to other people. Third, the examination of the use of products each knitting simulation knit found that sweater and muffler are the most appropriate knit products. And H2 was the most suitable sample for sweater and muffler products. This study tried to prove that knit textile can be visually evaluated through a computer knitting simulation without the actual knitting and then to provide accurate data to related companies or the academic circle.

• Korean Journal of Human Ecology
• /
• v.21 no.2
• /
• pp.285-298
• /
• 2012

Considerable number of women are enjoying swimming, however, the chest-region of swimsuits including cup inserts are not stable during swimming. The purpose of this study was to explore alternate designs and methods of stabilizing the breast cup to the swimmer's body by using 3D technology to design and position the pad and cup pattern of the swimming suit. To position the pad optimally, a 3D pattern of a nude woman was divided into blocks and different reduction rates were applied to blocks around the outlines of the breast. Two types of 3D patterns were developed. The first, referred to as the 'basic 3D pattern' provided for the curved surface of the breast point to be maintained with the remaining slack distributed evenly along the neckline, armhole and side seam. The second, referred to as the '1/3 shoulder moved 3D pattern' ignored the curved nature of the breast point by overlapping, with the resulting position of the shoulder strip moved toward the center. Three women of corresponding size and body shape evaluated the two 3D pattern designs as well as the conventional 2D pattern style of swimsuit. Respondents rated the'the basic 3D pattern' design highest in terms of stability and easy of movement in the chest region.

• Science of Emotion and Sensibility
• /
• v.20 no.2
• /
• pp.59-72
• /
• 2017

In the flood of information, many consumers want to choose the style that is suitable for their sensibility, which is usefulness and need for personalized services have been steadily increasing. This study made a video of personalized internet shopping mall and then university students (N=170) who have been using the internet shopping mall were asked to experience this and the change in consumer response was measured. According to level of personalization, this study found difference of web-site evaluation, satisfaction/internet behavior and quality evaluations of products. With regard to preferred types of internet shopping malls and the number of access, the difference was investigated. The results are as follows. First, subjects who experienced internet shopping mall of active personalization showed higher level of active personalization. Level of passive personalization differed depending on preferred types of internet shopping malls, for example, people who prefer online apparel shopping mall were low but people who prefer complex big shopping mall and social commerce were high. Second, after experiencing internet shopping mall of active personalization, satisfaction/internet behavior and quality evaluations of products did not change but passive personalization decreased and active personalization and web-site evaluation increased. Third, the number of access to internet shopping mall positively correlated with satisfaction/internet behavior and web-site evaluation, on the other hand, active personalization negatively correlated with satisfaction/internet behavior and web-site evaluation.

• Korean Journal of Human Ecology
• /
• v.11 no.4
• /
• pp.367-380
• /
• 2002

This study tries to search and analyze empirically the clothing merchandise purchase motivation and behavior made by foreign visitors who bought clothes as tour memorials in South Korea. To make a empirical study we made survey research and evaluated 30 items for the survey of clothing purchase motivation and 25 items for the survey of the importance and satisfaction revealed at the time of purchase with the standard of Likert 5 point. We had chosen 614 sample participants including 176 English speakers, 211 Japanese speakers, and 277 Chinese speakers among those foreign tourists who had purchased clothing materials at the Dongdaemun Shopping Mall. The result of analysis is as follows : As the result of factor analysis about the clothing merchandise purchase, we had 6 factors. They are (1)impulse, (2)utility, (3)fashionableness, (4)tour commemoration (5)psychology (6)economy. As the result of cluster analysis based on the standard of selected 6 factors of clothing merchandise purchase motivation, we could divide three groups such as (1)the pursuit type of clothing merchandise purchase, (2)the pursuit type of recreation of feeling (3)the pursuit type of tour memorial. As the result of analysis on the items of purchased clothing merchandise, the highest frequency of purchase was made by T-shirt, then by the order of sweater, blouse, and aloha shirt, pants, and jacket. In regard to the importance of items related to clothing merchandise, the highest importance was made by design, then by the order of price, size, color and utility, and style. In regard to the items related to the service of shopping mall, the highest satisfaction was made by the convenience of business hour. Meanwhile, the lowest satisfaction was made by the inconvenience of fitting room, the speaking and by ability of foreign languages by sales agents.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.30 no.2 s.150
• /
• pp.306-315
• /
• 2006

Today, the uniform is perceived being important, because it transmits the unique company image through visualization of images, improves efficiency of work, and increases the feeling of belonging. Therefore, the purpose of this project was to make the winter uniform for the personnel in the manufacturing and the business departments of Korea Tobacco and Ginseng Corporation. The functional, aesthetic, and symbolic aspects were fully considered throughout the design process for the new uniform. To reflect the worker's opinions, the questionnaire method was used. The questionnaire was made up of the questions about the evaluation criteria and design preference for uniform. The result of the survey was reflected on the uniform design. The uniform consists of a jumper and a pair of pants. The jumper for manufacturing workers has two patched pockets with gussets. It has a convertible collar and a horizontal tuck on front panel. It has an open fastener on the front and a flap to cover the fastener. It has two types of pockets inside for functional reason. The welt one is for a wallet and the small one is for a pen. For the material, melanges with two tones of gray color was used, so it gave the sophisticated look on the uniform. The jumper for business workers also has a convertible collar and inside pockets. It has two slant slash pockets with welt. In addition, it has an attachable fur collar which gives thermal protection and aesthetic effect. The fabric has small check pattern and was made of wool. The pants are a basic trouser style with two pleats on the front.

• The Research Journal of the Costume Culture
• /
• v.19 no.2
• /
• pp.346-359
• /
• 2011

The purpose of this study was to examine students' reasons and standards for purchasing their casual upper garments through online shoppin. A questionnaire composed of 26 items in five-point Likert type (14 items for measuring reasons, and 12 items for measuring standards for their purchase) was administered. The subjects were 422 male and female students attending middle schools, high schools and colleges located in the metropolitan region of Daegu. For a statistical analysis, a $3{\times}2$ two way ANOVA design (3 levels of schools: middle school, high school and college and 2 sexes: male and female) was involved, and Turkey's HFD multiple comparisons were made. The results showed that the reasons for students' purchasing casual upper garments through online shopping malls were as follows: quality, availability of discount coupons and points, other benefits such as special promotions, gifts, the easy return of goods and refunds, and also no trial of wearing shirts even at off-line stores. Significant differences in purchasing reasons through online shopping were found among middle schoo, high school and college students. There were also significant differences between male and female students mostly found in three variations of purchasing reasons: ease of availability of garments in contemporary fashion, convenient shopping without any restriction on time, and the decision to purchase with help from consumers' recommendations. Significant differences among middle school, high school and college students were found in the following standards for the students' decision to purchase their casual upper garments through online shopping: affordability, color, design, style, payment safety, and ease of maintenance (cleaning and ironing). Differences between male and female students were found to be significant in the following categories: affordability, fashionable, brand name, free delivery, product quality, coordination with other clothes, and consumers' recommendations on the products concerned.

• Journal of Korean Ophthalmic Optics Society
• /
• v.3 no.1
• /
• pp.39-73
• /
• 1998

In the 60's to 70's, frame selection was a purely mechanical consideration almost entirely dominated by the optician and his subjective judgments. Toddy the cosmetic factor is predominant. The variety of frame color, materials, and style means that the main burden of selection has passed to the customer, leaving the optician control a subtle factor often difficult to exert. Common materials include nickel silver, Hi-nickel alloy, bronze, stainless steel, gold, gold plated, gold clad, copper beryllium, titanium and sometimes aluminium. In manufacturing of metal frames with the materials, even though fashion is the stimulant of consumer demand, all the metal frame for the prescription to refractive correction should be designed correctly styled eyewear which is cosmetically pleasing, functional correct, physically comfortable, mechanically and structurally perfectible, and temperamentally acceptable.

• Journal of Preventive Medicine and Public Health
• /
• v.39 no.3
• /
• pp.221-228
• /
• 2006

Objectives: The purpose of this study was to shed further light on the effect of modifiable health behavior risk factors on dependence in activities of daily living, defined in a multidimensional fashion. Methods: The study participants were 10,278 middle aged Americans in a longitudinal health study, the Health and Retirement Survey (HRS). A multi-stage probability sampling design incorporating the effect of population sizes (Metropolitan and non-metropolitan), ethnicity (the non- Hispanic White, the Hispanic, and the Black), and age (age 51-61) was utilized. Basic Activities of Daily Living (ADL) were measured using five activities necessary for survival (impairment in dressing, eating, bathing, sleeping, and moving across indoor spaces). Explanatory variables were four health behavior risk factors included smoking, exercise, Body Mass Index (BMI), and alcohol consumption. Results: Most participants at baseline were ADL independent (1992). 97.8% of participants were independent in all ADL's at baseline and 78.2% were married. Approximately 27.5% were current smokers at baseline, and the subjects reported moderate or heavy exercise were 74.8%. All demographic characteristics and behavioral risk factors were significantly associated with the ADL status at Wave 4 except alcohol consumption. Risk behaviors such as current smoking, sedentary life style and high BMI at Wave 1 were associated with ADL status deterioration; however, moderate alcohol consumption tended to be more related to better ADL status than abstaining at Wave 4. ADL status at Wave 1 was the strongest factor and the next was exercise and smoking affecting ADL status at Wave 4. People who were in ADL dependent at Wave 1 were 15.17 times more likely to be ADL dependent at Wave 4 than people who were in ADL independent at Wave 1. Concerning smoking cigarettes, people who kept only light exercise or sedentary life style at Wave 1 were 1.70 times more likely to be died at Wave 4 than the people who did not smoke at Wave 1. Conclusions: All demographics and health behaviors at wave 1 had consistently similar OR trends for ADL status to each other except alcohol consumption. Smoking and exercise in health behaviors, and age and gender in demographics at Wave 1 were significant factors associated with ADL group separation at Wave 4.