• Korean Journal of Agricultural and Forest Meteorology
• /
• v.24 no.4
• /
• pp.295-304
• /
• 2022

The purpose of this study is to detect the sorghum panicle using YOLOv5 based on RGB images acquired by a unmanned aerial vehicle (UAV) system. The high-resolution images acquired using the RGB camera mounted in the UAV on September 2, 2022 were split into 512×512 size for YOLOv5 analysis. Sorghum panicles were labeled as bounding boxes in the split image. 2,000images of 512×512 size were divided at a ratio of 6:2:2 and used to train, validate, and test the YOLOv5 model, respectively. When learning with YOLOv5s, which has the fewest parameters among YOLOv5 models, sorghum panicles were detected with mAP@50=0.845. In YOLOv5m with more parameters, sorghum panicles could be detected with mAP@50=0.844. Although the performance of the two models is similar, YOLOv5s ( 4 hours 35 minutes) has a faster training time than YOLOv5m (5 hours 15 minutes). Therefore, in terms of time cost, developing the YOLOv5s model was considered more efficient for detecting sorghum panicles. As an important step in predicting sorghum yield, a technique for detecting sorghum panicles using high-resolution RGB images and the YOLOv5 model was presented.

• Korean Journal of Mineralogy and Petrology
• /
• v.35 no.3
• /
• pp.367-376
• /
• 2022

This study is a preliminary step to understand the reaction between various liquids and zeolite in the subduction zone environment. Stilbite, NaCa₄(Al₉Si₂₇)O₇₂·28(H₂O), was selected and high pressure study was conducted on compressional behavior by the pressure-transmitting medium (PTM). Water and NaHCO₃ solution that can exist in the subduction zone was used as PTM, and samples were pressurized from ambient to a maximum of 2.5 GPa. Below 1.0 GPa, both experiments show a low linear compressibility in the range of 0.001 to 0.004 GPa^-1 and a high bulk modulus of 220(1) GPa. This is presumably because the structure of the stilbite becomes very dense due to insertion of water molecules or cations into the channel. On the other hand, at 1.0 GPa or higher, the trends of the two experiments are different. In the water run, the linear compressibility of the c-axis is increased to 0.006(1) GPa^-1. In the NaHCO₃ run, the linear compressibility of the b- and c-axis is increased to 0.006(1) GPa^-1. The bulk modulus after 1.0 GPa shows values of 40(1) and 52(7) GPa in water and NaHCO₃ run, respectively, confirming that stilbite becomes more compressible than that before 1.0 GPa. It is caused by the migration of cations and water molecules inside the channel, as the water molecules in the PTM start to freeze and stop to insert toward the channel at 1.0 GPa or more. In the NaHCO₃ run, it is assumed that the distribution of extra-framework species inside the structure is changed by substitution of the Na⁺ cation. It can be expected from tendency of the relative intensity ratio of the (001) and (020) peaks which show a different from that of the water run.

• Journal of Service Research and Studies
• /
• v.13 no.1
• /
• pp.186-200
• /
• 2023

This study aims to improve the usability of the white cane, which is walking equipment that most local visually impaired people use and carry when going out, and to contribute to the prevention of safety accidents and the walking rights of visually impaired people by providing improvement and resolution measures for the problems identified. Also, this study is a study on the visually impaired, primarily targeting the 1st to 2nd degree visually impaired people, who cannot go out on their own without walking equipment such as a white cane, corresponding to 20% among approximately 250,000 blind and low vision people in the Korean population. In the study process, the concept has been developed from the user's point of view in order that the white cane becomes a real help in the walking step of the visually impaired and the improvement of usability of the white cane, the main walking equipment for the visually impaired, are done by problem identification through the Double Diamond Model of Design Thinking (Empathize → Define → Ideate → Prototype → Test (verify)). As a result of the investigation in the process of Empathy, a total of five issues was synthesized, including an increase in the proportion of the visually impaired people, an insufficient workforce situation to help all the visually impaired, an improvement and advancement of assistive devices essential for the visually impaired, problems of damage, illegal occupation, demolition, maintenance about braille blocks, making braille block paradigms for the visually impaired and for everyone. In Ideate and Prototype steps, situations derived from brainstorming were grouped and the relationship were made through the KJ method, and specific situations and major causes were organized to establish the direction of the concept. The derived solutions and major functions are defined in four categories, and representative situations requiring solutions and major functions are organized into two user scenarios. Ideas were visualized by arranging the virtual Persona and Customer Journey Map according to the situation and producing a prototype through 3D modeling. Finally, in the evaluation, the final concept derived is a device such a smart cane for guidance for the visually impaired as ① a smart cane emphasizing portability + ② compatibility with other electronic devices + ③ a product with safety and convenience.

• Korean Journal of Ecology and Environment
• /
• v.57 no.2
• /
• pp.102-110
• /
• 2024

This study presents a fine scale distribution of the endangered species, Odontobutis obscura, through field surveys and literature reviews. Using the mark-recapture method, the population size in major habitats was determined. Field surveys conducted on 18 streams in Geoje Island revealed that the O. obscura was only found in the main streams and tributaries of the Sanyang, Gucheon, and Buchun Streams, which are part of the Sanyang Stream watershed. The O. obscura exhibited relative abundances ranging from 0.5% to 35.3% at different locations, with certain spots showing higher relative abundances (18.8% to 35.3%), indicating major habitat areas. A review of six literature studies confirmed the presence of the O. obscura, although there were differences in occurrence status depending on the purpose, scope, and duration of the studies. Combining the results of field and literature surveys, it was found that the O. obscura inhabits the main and tributary streams of the Sanyang, Gucheon, and Buchun Streams, from the upper to lower reaches. Currently, the O. obscura population in the Sanyang Stream watershed maintains a stable habitat, but its limited distribution range suggests potential issues such as genetic diversity deficiency in the long term. The population size of the O. obscura was confirmed at two specific locations, with densities of 0.5 to 1.5 individuals per m². The average movement distance from the release point was 13.1 m, indicating the limited mobility characteristic of ambush predators. Understanding the distribution and population size of endangered species is the first step towards their conservation and protection. Based on this information, further research could significantly contribute to the conservation of the O. obscura.

• 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.

• Clinical and Experimental Reproductive Medicine
• /
• v.23 no.3
• /
• pp.247-268
• /
• 1996

It has been known for a long time that gonadotropins and steroid hormones play a pivotal role in a series of reproductive biological phenomena including the maturation of ovarian follicles and oocytes, ovulation and implantation, maintenance of pregnancy and fetal growth & development, parturition and mammary development and lactation. Recent investigations, however, have elucidated that in addition to these classic hormones, multiple growth factors also are involved in these phenomena. Most growth factors in reproductive organs mediate the actions of gonadotropins and steroid hormones or synergize with them in an autocrine/paracrine manner. The insulin-like growth factor(IGF) system, which is one of the most actively investigated areas lately in the reproductive organs, has been found to have important roles in a wide gamut of reproductive phenomena. In the present communication, published literature pertaining to the intrauterine IGF system will be reviewed preceded by general information of the IGF system. The IGF family comprises of IGF-I & IGF-II ligands, two types of IGF receptors and six classes of IGF-binding proteins(IGFBPs) that are known to date. IGF-I and IGF-II peptides, which are structurally homologous to proinsulin, possess the insulin-like activity including the stimulatory effect of glucose and amino acid transport. Besides, IGFs as mitogens stimulate cell division, and also play a role in cellular differentiation and functions in a variety of cell lines. IGFs are expressed mainly in the liver and messenchymal cells, and act on almost all types of tissues in an autocrine/paracrine as well as endocrine mode. There are two types of IGF receptors. Type I IGF receptors, which are tyrosine kinase receptors having high-affinity for IGF-I and IGF-II, mediate almost all the IGF actions that are described above. Type II IGF receptors or IGF-II/mannose-6-phosphate receptors have two distinct binding sites; the IGF-II binding site exhibits a high affinity only for IGF-II. The principal role of the type II IGF receptor is to destroy IGF-II by targeting the ligand to the lysosome. IGFs in biological fluids are mostly bound to IGFBP. IGFBPs, in general, are IGF storage/carrier proteins or modulators of IGF actions; however, as for distinct roles for individual IGFBPs, only limited information is available. IGFBPs inhibit IGF actions under most in vitro situations, seemingly because affinities of IGFBPs for IGFs are greater than those of IGF receptors. How IGF is released from IGFBP to reach IGF receptors is not known; however, various IGFBP protease activities that are present in blood and interstitial fluids are believed to play an important role in the process of IGF release from the IGFBP. According to latest reports, there is evidence that under certain in vitro circumstances, IGFBP-1, -3, -5 have their own biological activities independent of the IGF. This may add another dimension of complexity of the already complicated IGF system. Messenger ribonucleic acids and proteins of the IGF family members are expressed in the uterine tissue and conceptus of the primates, rodents and farm animals to play important roles in growth and development of the uterus and fetus. Expression of the uterine IGF system is regulated by gonadal hormones and local regulatory substances with temporal and spatial specificities. Locally expressed IGFs and IGFBPs act on the uterine tissue in an autocrine/paracrine manner, or are secreted into the uterine lumen to participate in conceptus growth and development. Conceptus also expresses the IGF system beginning from the peri-implantation period. When an IGF family member is expressed in the conceptus, however, is determined by the presence or absence of maternally inherited mRNAs, genetic programming of the conceptus itself and an interaction with the maternal tissue. The site of IGF action also follows temporal (physiological status) and spatial specificities. These facts that expression of the IGF system is temporally and spatially regulated support indirectly a hypothesis that IGFs play a role in conceptus growth and development. Uterine and conceptus-derived IGFs stimulate cell division and differentiation, glucose and amino acid transport, general protein synthesis and the biosynthesis of mammotropic hormones including placental lactogen and prolactin, and also play a role in steroidogenesis. The suggested role for IGFs in conceptus growth and development has been proven by the result of IGF-I, IGF-II or IGF receptor gene disruption(targeting) of murine embryos by the homologous recombination technique. Mice carrying a null mutation for IGF-I and/or IGF-II or type I IGF receptor undergo delayed prenatal and postnatal growth and development with 30-60% normal weights at birth. Moreover, mice lacking the type I IGF receptor or IGF-I plus IGF-II die soon after birth. Intrauterine IGFBPs generally are believed to sequester IGF ligands within the uterus or to play a role of negative regulators of IGF actions by inhibiting IGF binding to cognate receptors. However, when it is taken into account that IGFBP-1 is expressed and secreted in primate uteri in amounts assessedly far exceeding those of local IGFs and that IGFBP-1 is one of the major secretory proteins of the primate decidua, the possibility that this IGFBP may have its own biological activity independent of IGF cannot be excluded. Evidently, elucidating the exact role of each IGFBP is an essential step into understanding the whole IGF system. As such, further research in this area is awaited with a lot of anticipation and attention.

• Journal of Korean Society of Forest Science
• /
• v.25 no.1
• /
• pp.13-47
• /
• 1975

Among the many species of bamboo, it is well known that the dwarf-type is widely distributed in the tropical regions, and the slender type in temperated zone. In the temperated zone the trees have extensively differentiated into one hundred species in 50 genera. In many oriental countries, the bamboo wood is being used as a material for construction and for the manufacture of technical instruments. The bamboo shoot is also regarded as a good and delicious edible resource. Moreover, recent medical investigation verifies that the sap of certain species of the bamboo is an antibiotic effect against cancer. Fortunately, it is very easy to propagate the bamboo trees by using cutting from southeastern Asian countries. This important resource can further be used as a significant source of pulp, which is becoming increasingly important. The classification system of this significant resource has not been completely established to date, even though its importance has been emphasized. Initiated by Canlevon Linne in the 18th century, a classification method concerning the morphological characteristics of flowers was the first step in developing a classification. But it was not an easy task to accomplish, because this type of classification system is based on the sexual organs in bamboo trees. Because the bamboo has a long life cycle of 60-120 years and classification according to this method was very difficult as the materials for the classification are not abundant and some species have changed, even though many references related to the morphological classification of bamboo trees are available nowadays. So, the certification of bamboo trees according to the morphological classification system is not reasonable for us. Consequently, the classification system of bamboo trees on the basis of endomorphological characteristics was initiated by Chinese-born Liese. And classification method based on the morphological characteristics of the vascular bundle was developed by Grosser. These classification methods are fundamentally related to Holltum's classification method, which stressed the morphology of the ovary. The author investigated to re-establish a new classification method based on the vascular sheath. Twenty-six species in 11 genera which originated from Formosa where used in the study. The results obtained from the investigation were somewhat coordinated with those of Crosser. Many difficulties were found in distinguishing the species of Bambusa and Dendrocalamus. These two species were critically differentiated under the new classification system, which is based on the existence of a separated vascular bundle sheath in the bamboo. According to these results, it is recommended that Babusa divided into two groups by placing it into either subspecies or the lower categories. This recommendation is supported by the observation that the evolutional pattern of the bamboo thunk which is from outward to inward. It is also supported by the viewpoint that the fundamental hypothesis in evolution is from simple to complex. There remained many problems to be solved through more critical examination by comparing the results to those of the classification based on the sexual organs method. The author observed the figure of the cross-sectional area of vascular trunk of bamboo tree and compared the results with those of Grosser and Liese, i.e. A, $B_1$, $B_2$, C, and D groups in classification. Group A and $B_2$ were in accordance with the results of those scholars, while group D showed many differences, Grosser and Liese divided bamboo into "g" type and "h" type according to the vascular bundle type; and they included Dendrocalamus and Bambusa in Group D without considering the type of vascular bundle sheath. However, the results obtained by the author showed that Dendrocalamus and Bambusa are differentiated from each other. By considering another group, "i" identified according to the existence of separated vascular bundle sheath. Bambusa showed to have a separated vascular bundle sheath while Dendrocalamus does not have a separated vascular bundle sheath. Moreover, Bambusa showed peculiar characteristics in the figure of vascular development, i.e., one with an inward vascular bundle sheath and the other with a bivascular bundle sheath (inward and outward). In conclusion, the bamboo species used in this experiment were classified in group D, without any separated vascular bundle sheath, and in group E, with a vascular bundle sheath. Group E was divided into two groups, i.e., and group $E_1$, with bivascular sheath, and group $E_2$, with only an inward vascular sheath. Therefore, the Bambusa in group D as described by Grosser and Liese was included in group E. Dendrocalamus seemed to be the middle group between group $E_l$ and group $E_2$ under this classification system which is summarized as follows: Phyllostachys-type: Group A - Phyllostachys, Chymonobambus, Arundinaria, Pseudosasa, Pleioblastus, Yashania Pome-type: Group $B_2$ - Schizostachyum, Melocanna Hemp-type: Group D - Dendrocalamu Bambu-type: Group $E_1$ - Bambusa ghi.

• Journal of Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.45-69
• /
• 2016

Recently, sentiment analysis using open Internet data is actively performed for various purposes. As online Internet communication channels become popular, companies try to capture public sentiment of them from online open information sources. This research is conducted for the purpose of analyzing pulbic sentiment of Korean Top-10 companies using a multi-categorical sentiment lexicon. Whereas existing researches related to public sentiment measurement based on big data approach classify sentiment into dimensions, this research classifies public sentiment into multiple categories. Dimensional sentiment structure has been commonly applied in sentiment analysis of various applications, because it is academically proven, and has a clear advantage of capturing degree of sentiment and interrelation of each dimension. However, the dimensional structure is not effective when measuring public sentiment because human sentiment is too complex to be divided into few dimensions. In addition, special training is needed for ordinary people to express their feeling into dimensional structure. People do not divide their sentiment into dimensions, nor do they need psychological training when they feel. People would not express their feeling in the way of dimensional structure like positive/negative or active/passive; rather they express theirs in the way of categorical sentiment like sadness, rage, happiness and so on. That is, categorial approach of sentiment analysis is more natural than dimensional approach. Accordingly, this research suggests multi-categorical sentiment structure as an alternative way to measure social sentiment from the point of the public. Multi-categorical sentiment structure classifies sentiments following the way that ordinary people do although there are possibility to contain some subjectiveness. In this research, nine categories: 'Sadness', 'Anger', 'Happiness', 'Disgust', 'Surprise', 'Fear', 'Interest', 'Boredom' and 'Pain' are used as multi-categorical sentiment structure. To capture public sentiment of Korean Top-10 companies, Internet news data of the companies are collected over the past 25 months from a representative Korean portal site. Based on the sentiment words extracted from previous researches, we have created a sentiment lexicon, and analyzed the frequency of the words coming up within the news data. The frequency of each sentiment category was calculated as a ratio out of the total sentiment words to make ranks of distributions. Sentiment comparison among top-4 companies, which are 'Samsung', 'Hyundai', 'SK', and 'LG', were separately visualized. As a next step, the research tested hypothesis to prove the usefulness of the multi-categorical sentiment lexicon. It tested how effective categorial sentiment can be used as relative comparison index in cross sectional and time series analysis. To test the effectiveness of the sentiment lexicon as cross sectional comparison index, pair-wise t-test and Duncan test were conducted. Two pairs of companies, 'Samsung' and 'Hanjin', 'SK' and 'Hanjin' were chosen to compare whether each categorical sentiment is significantly different in pair-wise t-test. Since category 'Sadness' has the largest vocabularies, it is chosen to figure out whether the subgroups of the companies are significantly different in Duncan test. It is proved that five sentiment categories of Samsung and Hanjin and four sentiment categories of SK and Hanjin are different significantly. In category 'Sadness', it has been figured out that there were six subgroups that are significantly different. To test the effectiveness of the sentiment lexicon as time series comparison index, 'nut rage' incident of Hanjin is selected as an example case. Term frequency of sentiment words of the month when the incident happened and term frequency of the one month before the event are compared. Sentiment categories was redivided into positive/negative sentiment, and it is tried to figure out whether the event actually has some negative impact on public sentiment of the company. The difference in each category was visualized, moreover the variation of word list of sentiment 'Rage' was shown to be more concrete. As a result, there was huge before-and-after difference of sentiment that ordinary people feel to the company. Both hypotheses have turned out to be statistically significant, and therefore sentiment analysis in business area using multi-categorical sentiment lexicons has persuasive power. This research implies that categorical sentiment analysis can be used as an alternative method to supplement dimensional sentiment analysis when figuring out public sentiment in business environment.

• Restorative Dentistry and Endodontics
• /
• v.24 no.1
• /
• pp.166-179
• /
• 1999

Apical sealing is essential for the success of surgical endodontic treatment. Root-end cavity is apt to be contaminated with moisture or blood, and is not always easy to be dried completely. The purpose of this study was to evaluate the influence of dry methods of retrocavity on the apical seal in endodontic surgery. Apical seal was investigated through the evaluation of apical leakage and adaptation of filling material over the cavity wall. To investigate the influence of various dry methods on the apical leakage, 125 palatal roots of extracted human maxillary molar teeth were used. The clinical crown of each tooth was removed at 10 mm from the root apex using a slow-speed diamond saw and water spray. Root canals of the all the specimens were prepared with step-back technique and filled with gutta-percha by lateral condensation method. After removing of the coronal 2 mm of filling material, the access cavities were closed with Cavit$^{(R)}$. Two coats of nail polish were applied to the external surface of each root. Apical three millimeters of each root was resected perpendicular to the long axis of the root with a diamond saw. Class I retrograde cavities were prepared with ultrasonic instruments. Retrocavities were washed with physiologic saline solution and dried with various methods or contaminated with human blood. Retrocavities were filled either with IRM, Super EBA or composite resin. All the specimens were immersed in 2% methylene blue solution for 7 days in an incubator at $37^{\circ}C$. The teeth were dissolved in 14 ml of 35% nitric acid solution and the dye present within the root canal system was returned to solution. The leakage of dye was quantitatively measured via spectrophotometric method. The obtained data were analysed statistically using one-way ANOVA and Duncan's Multiple Range Test. To evaluate the influence of various dry methods on the adaptation of filling material over the cavity wall, 12 palatal roots of extracted human maxillary molar teeth were used. After all the roots were prepared and filled, and retrograde cavities were made and filled as above, roots were sectioned longitudinally. Filling-dentin interface of cut surfaces were examined by scanning electron microscope. The results were as follows: 1. Cavities dried with paper point or compressed air showed less leakage than those dried with cotton pellet in Super EBA filled cavity (p<0.05). However, there was no difference between paper point- and compressed air-dried cavities. 2. When cavities were dried with compressed air, dentin-bonded composite resin-filled cavities showed less apical leakage than IRM- or Super EBA-filled ones (p<0.05). 3. Regardless of the filling material, cavities contaminated with human blood showed significantly more apical leakage than those dried with compressed air after saline irrigation (p<0.05). 4. Outer half of the cavity showed larger dentin-filling interface gap than inner half did when cavities were filled with IRM or Super EBA. 5. In all the filling material groups, cavities contaminated with blood or dried with cotton pellets only showed larger defects at the base of the cavity than ones dried with paper points or compressed air.

• 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.