• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.

• Journal of Trauma and Injury
• /
• v.19 no.2
• /
• pp.97-104
• /
• 2006

Purpose: Many factors influence the occurrence and severity of geriatric trauma, and regional weather is regarded as one factor that influences geriatric trauma. In this study, to predict the type, severity, and incidence of geriatric trauma patient, we analyzed the influence of regional weather on geriatric trauma. Methods: The subjects of our investigation were trauma patients over sixty-five years of age who visited the Emergency Department (ED) of Yeungnam University Hospital during a one-year period. We retrospectively reviewed the medical charts of 436 geriatric trauma patients, and the data were analyzed by using SPSS 12.0 for Window. The weather was based on data from the Korea Meteorological Administration. Results: The average age was 72.8 years old, and the ratio of males to females was 1:1.1. The mean spell out ISS was 10.8, and no difference was found between males and females. Slips were the most common cause of trauma. The largest numbers of aged trauma patients, 46, visited the ED in May, and the smallest number of such patients, 24, visited the ED in December. In addition to, summer saw the largest number of aged trauma patients. The type of trauma, the Injury Severity Score, and the number of patients had no relationship with season. On sunny days, the ISS was larger in patients who had hypotension and who had tachycardia. On rainy day, the ISS was larger in male patients and cultivator accident patients. The number of patients was larger on partly cloudy days. Conclusion: In spring and summer and on partly cloudy days, we must be prepared to treat aged traumatized patients in the E.D. On rainy days, visual sensation, tactual sense, and acoustic sense must be closely examined. In addition,on rainy day, aged male traumatized patients or cultivator accident patients must to be closely observation.

• Journal of Korean Society of Occupational and Environmental Hygiene
• /
• v.24 no.4
• /
• pp.547-555
• /
• 2014

Objective: This study was conducted to investigate the association between subjective distress symptoms and argon welding among workers in Gyeongnam Province shipyard. Method: 31 argon and 29 non-argon welding workers were selected as study subjects in order to measure concentrations of personal dust, welding fumes and other hazardous materials such as ZnO, Pb, Cr, FeO, MnO, Cu, Ni, $TiO_2$, MgO, NO, $NO_2$, $O_3$, $O_2$, $CO_2$, CO and Ar. An interviewer-administered questionnaire survey was also performed on the same subjects. The items queried were as follows: age, height, weight, working duration, welding time, welding rod amounts used, drinking, smoking, and rate of subjective distress symptoms including headache and other symptoms such as fever, vomiting and nausea, metal fume fever, dizziness, tingling sensations, difficulty in breathing, memory loss, sleep disorders, emotional disturbance, hearing loss, hand tremors, visual impairment, neural abnormality, allergic reaction, runny nose and stuffiness, rhinitis, and suffocation. Statistical analysis was performed using SPSS software, version 18. Data are expressed as the mean ${\pm}SD$. An ${\chi}^2$-test and a normality test using a Shapiro wilk test were performed for the above variables. Logistic regression analysis was also conducted to identify the factors that affect the total score for subjective distress symptoms. Result: An association was shown between welding type (argon or non-argon welding) and the total score for subjective distress symptoms. Among the rate of complaining of subjective distress symptoms, vomiting and nausea, difficulty breathing, and allergic reactions were all significantly higher in the argon welding group. Only the concentration of dust and welding fumes was shown to be distributed normally after natural log transformation. According to logistic regression analysis, the correlations of working duration and welding type (argon or non-argon) between the total score of subjective distress symptoms were found to be statistically significant (p=0.041, p=0.049, respectively). Conclusion: Our results suggest that argon welding could cause subjective distress symptoms in shipyard workers.

• Journal of Korean Ophthalmic Optics Society
• /
• v.18 no.1
• /
• pp.61-65
• /
• 2013

Purpose: This study was performed to identify the relations between glaucoma and dietary patterns in Korean adults, aged 30 years or older. Methods: 6,453 subjects (male 2,759, female 3,694 subjects) who aged 30 years or older from the Korea National Health and Nutrition Examination Survey were included. Glaucoma was assessed based on the subjects were diagnosed by a doctor. At first, waist circumference and body mass index (BMI) were compared to investigate the association between glaucoma and obesity index. Secondly, dietary patterns using factor analysis for external blue colored fish(mackerel), seaweed(seaweed, laver), bean(bean, soybean milk), vegetables(sprout, spin, carrot, cabbage, cucumber, pepper, tomato, pumpkin) and fruits(tangerine, apple, pear, strawberry, grape, peach) were obtained. Lastly, regression analysis was used for glaucoma and emerged dietary factors. Results: In glaucoma and normal group, waist circumference and BMI were $82.25{\pm}9.96cm$ and $23.27{\pm}2.84kg/m^2$, $75.79{\pm}13.79cm$ and $22.39{\pm}4.06kg/m^2$, respectively. There was statistically significant difference (p<0.05). Three dietary patterns were emerged from the factor analysis, and factor 1 was included sprout, spin, carrot, cabbage, seaweed, cucumber, mackerel, laver. The factor 1 of dietary pattern was inversely related to the prevalence of glaucoma (OR: 0.697, 95%CI: 0.496-0.980, p=0.038), and model adjusting for age and sex was showed the same trend (OR: 0.698, 95%CI: 0.497-0.982, p=0.039). Conclusions: This study was analysed first with glaucoma and dietary pattern in Korean, and vegetables, laver and mackerel significant associated to reduce the risk of glaucoma.

• Spatial Information Research
• /
• v.11 no.1
• /
• pp.41-60
• /
• 2003

It is noted that the paper mapping system for University Building Facility has many limitations in terms of data maintenance, real-time GIS data acquisition, and economic efficiency. The aim of this research was to evaluate an operational potential of an on site real-time mobile GIS technique to resolve the problem faced in the university. The idea is based upon the recent trends in the field of 'Telecommunication and Information Technology' that uses a PDA (personal Digital Assistants), wireless network computing, mobile computing, etc. A real time mobile GIS approach has been adopted, in which a PDA is linked to a wireless internet and field workers record data on the computer at the site and analyse data on site. While there should be a considerable number and variety of factors associated with real-time mobile GIS quality, this research focuses on three criteria that are identified as fundamental to customer requirements; (1) data quarry (2) spatial analysis (3) real-time GIS database building. 'Art--empirical study for a case study facility has been conducted to confirm the validity for the system. The system has been checked experimentally, enabled the field users to quarry the data required simply and execute spatial analysis (buffer, overlay etc.,) accommodating versatile alternatives on the site. Detailed visual maps can be generated over large areas quickly and easily. The PDA interface, in particular, were ideally suited for field users to interactively displaying positional information with attribute data. This system has shown to be quite convenient to maintaining a highly reliable database since it could playa crucial role in documenting at real-time basis temporal and spatial changes occurred in the facilities. It is anticipated that this research output will greatly serve to introduce the reliable and cost-effective facility mapping system in the university by overcoming serious constraints suffered from the past non-real time mobile GIS approach.

• School Mathematics
• /
• v.19 no.2
• /
• pp.229-248
• /
• 2017

This study investigated the factors that should be considered when teaching proportional expression and proportional distribution through literature review. Based on these results, we analyzed and compared Korean and foreign mathematics textbooks on proportional expression and proportional distribution longitudinally and horizontally to search for desirable methods of organizing the unit of proportional expression and proportional distribution in mathematics textbooks. For longitudinal analysis, we took the mathematics textbooks according to the national curriculum since the 5th one. For horizontal analysis, we selected the mathematics textbooks of Japan, Singapore, and China. In each textbook, the contents and the order in relation to proportional expression and proportional distribution, the definitions of terminology, and the contexts and the visual representations for introducing related concepts are selected as the analysis framework. The results of analysis revealed many characteristics and the differences in ways of dealing contents about proportional expression and proportional distribution. Based on these results, we suggested some implications for writing the unit of proportional expression and proportional distribution in elementary mathematics textbooks.

• Korean Journal of Community Nutrition
• /
• v.25 no.3
• /
• pp.189-203
• /
• 2020

Objectives: This study was performed to identify the current barriers of obesity management for children using Community Child Care Centers and their caregivers (parents and teachers working in the Centers). Further, this study explored the possibility of utilizing a mobile phone application for tailored obesity prevention and management programs to overcome the current difficulties associated with children's obesity management. Methods: The qualitative data were collected through in-depth interviews with 20 obese and overweight children or children who wanted to participate in this study using Community Child Care Centers, 12 teachers working at the Centers, and a focus group interview with five parents of children using the Centers. Data were analyzed with a thematic approach categorizing themes and sub-themes based on the transcripts. Results: The current barriers of obesity management of obese and overweight children using Community Child Care Centers were lack of self-directed motivation regarding obesity management (chronic obesity-induced lifestyles and reduced self-confidence due to stigma) and lack of support from households and Community Child Care Centers (latchkey child, inconsistency in dietary guidance between the Center and household, repetitive pressure to eat, and absence of regular nutrition education). Mobile phone applications may have potential to overcome the current barriers by providing handy and interesting obesity management based on visual media (real-time tracking of lifestyles using behavior records and social support using gamification), environmental support (supplementation of parental care and network-based education between the Community Child Care Center and household), and individualized intervention (encouragement of tailored and gradual changes in eating habits and tailored goal setting). It is predicted that the real-time mobile phone program will provide information for improving nutritional knowledge and behavioral skills as well as lead to sustainable children's coping strategies regarding obesity management. In addition, it is expected that environmental factors may be improved by network-based education between the Community Child Care Centers and households using the characteristics of mobile phones, which are free from space and time constraints. Conclusions: The tailored education program for children using Community Child Care Centers based on mobile phones may prevent and reduce childhood obesity by overcoming the current barriers of obesity management for children, providing environmental and individualized support to promote healthy lifestyles and quality of life in the future.

• Journal of The Korean Association For Science Education
• /
• v.37 no.1
• /
• pp.181-192
• /
• 2017

This study aims to examine the educational effects of the SSI program using Digital Storytelling (DST) approaches. Since DST provides students opportunities to express their own opinions in the form of stories and to share learning outcomes through the web, we developed and implemented SSI program by adopting the concept of DST in order to produce synergistic effects on student learning. Twenty-four 9th graders who enthusiastically engaged in the DST-based SSI program participated in this study. The students responded to focus group interviews after the instruction, and all interviews were transcribed for analysis. The results indicated that the students became aware of socio-ethical perspectives of each SSI topic while searching and collecting data by themselves. They also felt the necessity to consider multiple perspectives around the issues by having discussions with group members. Second, pre-producing DST allowed students to negotiate to settle on a group discussion, and to use emotional contents that can lead viewers to have sympathy. In addition, while producing DST, students considered various factors such as design, soundtrack, visual effects, and screen composition in order to express their opinions and convey their messages more effectively. In the stage of sharing DST outcomes and receiving feedback, they realized new perspectives that they did not perceive in the previous production process, and to move them into an action for resolving the problems caused by SSI. This study showed the potentials of DST-based SSI instruction as a good strategy to support students' SSI engagement.

• Journal of the Korean Society of Costume
• /
• v.66 no.6
• /
• pp.95-109
• /
• 2016

The process of making fashion accessories brings many benefits to visually impaired people. It helps the development of their tactile sensation, improves their aesthetic attitude, creativity, and imagination, and brings energy to their life. The current accessory making education in Korea is targeted to people with normal vision, and does not reflect the specific needs of the visually impaired. This causes a lot of trouble in educating the visually impaired. Therefore, this study aims to increase the efficiency of accessory making education by selecting curriculum and material that are appropriate and systematic for the visually impaired. In order to understand visual impairment, literature review was conducted to examine the definition and classifications of the impairment. Research was also done regarding material types and production method of fashion accessory. The study also designed an education program, including teaching guidelines, by referring to literature related to fashion accessory design, teaching guideline. The study conducted 12 experiments based on the results from April to June 2014. After revising and supplementing the teaching method, 2nd set of experiments were conducted from July to September 2014. Each class lasted for one and a half hour every week, and the class was composed of 6 students, 1 instructor and 1 volunteer. After each class, interviews were done to collect student opinions about the curriculum and material. Total experiment results and interview were reflected in constructing the final education program. The result of the experiments and interviews showed that selecting the accessory materials for class must be selected carefully. The visually-impaired students were very sensitive to the size, texture, shape, hole size of beads and length of material. After the experiments, all the students were able to distinguish the size and shape of the beads well, and finished the accessories using the given materials. The study verified that there is a need for another approach when educating the visually impaired about fashion accessory design. Many factors need to be considered from selecting materials to class contents. This study will contribute as a reference to fashion accessory making education as well as other design education for the visually impaired.

• The Journal of Korean Academy of Prosthodontics
• /
• v.41 no.5
• /
• pp.565-581
• /
• 2003

Statement of problem : A scientific examination and understanding of overall aspects of the natural dentition is the first step involved in making a satisfactory selection in the shade of an aesthetic prosthesis. Proper natural shade selection of the prosthetic restoration that is in harmony with the remaining dentition is as important aesthetically, as harmony of form and function in the anterior dentition. Clinically, the most commonly applied method of shade selection has been visual, but because of the subjective nature inherent to this method, shade selection results are variable and can be influenced by such factors as the technician, the type of shade guide used, and the type and intensity of the lighting. Purpose : The purpose of this study was to develop a more objective and scientific approach to examining and understanding the shade of teeth, which has in turn lead to the development of a number of shade analysis devices that present a more objective method of shade analysis. Material and Method : In this study, the shades of healthy anterior teeth were examined and analyzed using the recently developed digital shade analysis of the $ShadeScan^{TM}$ System. The study examined 80 individuals in their twenties, 40 males and 40 females, presenting 6 healthy, unrestored maxillary anterior teeth. Tooth brushing and oral prophylaxis were performed prior to evaluation. The ShadeScan handpiece was used to acquire images of the 6 maxillary anterior teeth. These images were analyzed using the Vita/Classical mode of the $ShadeScan^{TM}$ Software, and shade maps of each tooth were acquired and divided into cervical, middle, and incisal thirds. The shade distribution of each third, left and right symmetry, and gender differences were investigated and analyzed. Results : The results of the study are as follows : 1. An overwhelming majority of the examined teeth were found to possess shades belong to Group A, with the greatest variations occurring at the middle and cervical thirds of the maxillary central and lateral incisors, in both male and female subjects. 2. Canines of both male and female subjects showed left and right symmetry with uniform shade distribution of A4 and C4, while the lateral and central incisors showed left-right symmetry of the incisal 1/3 with a uniform shade distribution of A2 and A3 shades 3. No significant differences in shade distribution were seen between genders in maxillary canines, whereas maxillary central and lateral incisors showed differences at the middle and cervical thirds between male and female subjects The results of this study show that with the exception of maxillary canines, maxillary anterior teeth display a diverse shade distribution as well as gender differences. Conclusion : Clinically, when making a shade selection using the existing shade guide, one must consider the fact that even a single tooth consists of a variety of shades. The results of this study show that when selecting a shade from a number of groups is difficult, shades from A group are the most consistent with the natural shade or maxillary anterior teeth.