• 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 Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.113-127
• /
• 2016

Among the techniques for recommendation, collaborative filtering (CF) is commonly recognized to be the most effective for implementing recommender systems. Until now, CF has been popularly studied and adopted in both academic and real-world applications. The basic idea of CF is to create recommendation results by finding correlations between users of a recommendation system. CF system compares users based on how similar they are, and recommend products to users by using other like-minded people's results of evaluation for each product. Thus, it is very important to compute evaluation similarities among users in CF because the recommendation quality depends on it. Typical CF uses user's explicit numeric ratings of items (i.e. quantitative information) when computing the similarities among users in CF. In other words, user's numeric ratings have been a sole source of user preference information in traditional CF. However, user ratings are unable to fully reflect user's actual preferences from time to time. According to several studies, users may more actively accommodate recommendation of reliable others when purchasing goods. Thus, trust relationship can be regarded as the informative source for identifying user's preference with accuracy. Under this background, we propose a new hybrid recommender system that fuses CF and social network analysis (SNA). The proposed system adopts the recommendation algorithm that additionally reflect the result analyzed by SNA. In detail, our proposed system is based on conventional memory-based CF, but it is designed to use both user's numeric ratings and trust relationship information between users when calculating user similarities. For this, our system creates and uses not only user-item rating matrix, but also user-to-user trust network. As the methods for calculating user similarity between users, we proposed two alternatives - one is algorithm calculating the degree of similarity between users by utilizing in-degree and out-degree centrality, which are the indices representing the central location in the social network. We named these approaches as 'Trust CF - All' and 'Trust CF - Conditional'. The other alternative is the algorithm reflecting a neighbor's score higher when a target user trusts the neighbor directly or indirectly. The direct or indirect trust relationship can be identified by searching trust network of users. In this study, we call this approach 'Trust CF - Search'. To validate the applicability of the proposed system, we used experimental data provided by LibRec that crawled from the entire FilmTrust website. It consists of ratings of movies and trust relationship network indicating who to trust between users. The experimental system was implemented using Microsoft Visual Basic for Applications (VBA) and UCINET 6. To examine the effectiveness of the proposed system, we compared the performance of our proposed method with one of conventional CF system. The performances of recommender system were evaluated by using average MAE (mean absolute error). The analysis results confirmed that in case of applying without conditions the in-degree centrality index of trusted network of users(i.e. Trust CF - All), the accuracy (MAE = 0.565134) was lower than conventional CF (MAE = 0.564966). And, in case of applying the in-degree centrality index only to the users with the out-degree centrality above a certain threshold value(i.e. Trust CF - Conditional), the proposed system improved the accuracy a little (MAE = 0.564909) compared to traditional CF. However, the algorithm searching based on the trusted network of users (i.e. Trust CF - Search) was found to show the best performance (MAE = 0.564846). And the result from paired samples t-test presented that Trust CF - Search outperformed conventional CF with 10% statistical significance level. Our study sheds a light on the application of user's trust relationship network information for facilitating electronic commerce by recommending proper items to users.

• Journal of Korean Ophthalmic Optics Society
• /
• v.10 no.3
• /
• pp.205-213
• /
• 2005

To evaluate the reliability of binocular vision measurements by phorometry. 90 students volunteered to participate in this study. 25 subjects were males, and 65 were females, they ranged in ages from 21 to 30 years. All subjects had normal ocular and systematic health, and all of them had at least 1.0 visual acuity with their best correction. At negative relative convergence(NRC) measurement in distance, the percentage of subjects is included in expected value(blur point/break point/recovery point) was 78%/61%/67%, divergence excess(DE) was 9%/31%/33%, and divergence insufficience(DI) was 13%/8%/9%, respectively. And positive relative convergence(PRC) measurement, includes expected value was 20%/46%/39%, convergence excess(CE) was 22%/14%/16%, and convergence insufficience(CI) was 35%/40%/45%. AC/A ratios of 42 subjects were normal. 38 were low, and the rest of them high. A low AC/A ratio is usually the result of a small vergence response in relation to accommodation. Negative relative convergence(NRC) at near, includes expected value was 26%/29%/44%, divergence excess(DE) was 61%/33%/24%, and divergence insufficience(DI) was 3%/38%/32%. And PRC at near, includes expected value was 33%/40%/31%, convergence excess(CE) was 61%/23%/42%, and convergence insufficience(CI) was 6%/37%/27%. For the near point of convergence(NPC) test, 58% of their subjects had a break of ${\leq}8cm$ with the accommodative target. In case of NRA(PRA) measurement, the expected value was 41%(33%). Accommodative insufficiency (AI) was 33%(43%), and accommodative excess(AE) was 26%(24%), respectively. AE was related to respectively low values of NRA. AI and CE are associated with high value of NRA, and the dysfunction of convergence excess combined with AE was related to a normal-high values of NRA. PRA in AI was related to a low value, wheres the dysfunction are associated with high values of PRA.

• Journal of Korean Ophthalmic Optics Society
• /
• v.17 no.2
• /
• pp.185-194
• /
• 2012

Purpose: The changes of phoria and subjective asthenopia before and after viewing were compared based on 2D image and two ways of 3D images, and presented for references of 3D image watching and production. Methods: Change in phoria was measured before and after watching 2D image, 3D-FPR and 3D-SG images for 30 minutes with a target of 41 university students at 20-30 years old (male 26, female 15). Paired t-test and Pearson correlation between changed phoria and subjective symptoms which were measured using questionnaires were evaluated by before and after watching each images. Results: Right after watching 2D image, exophoria was increased by 0.5 $\Delta$, in distance and near, but it was not a significant level. Right after watching 3D image, exophoria was increased by 1.0~1.5 $\Delta$, and 1.5~2.0 $\Delta$, in distance and near, respectively when compared with before watching. In the significant level, exophoria tended to increase. Changes in near was increased more by 0.5 $\Delta$, compared with those in distance. Changes based on way of 3D-FPR and 3D-SG image were less than 0.5 $\Delta$, and there was almost no difference. In terms of visual subjective symptoms, eye strain was increased in 3D image compared with that in 2D image. In addition, there was no difference depending on way of image. In terms of Pearson correlation between phoria change and eye strain, as exophoria was increased, eye strain was increased. Conclusions: Watching 3D image increased eye strain compared with watching 2D image, and accordingly exophoria tended to increase.

• Cartoon and Animation Studies
• /
• s.44
• /
• pp.117-141
• /
• 2016

The various representations are receiving attention in modern society that has so many contents. Among them, the fantasy shows that you can not see in reality. But the intention of these fantasies is not giving a visual fun. The fantasy show the reality through stories that are not in reality. The fantasy that allows readers to continue to make the suspect between the real and the imagination and that suspect arises from the desire of real life. If the desire break the community, the social ideology will collapses. Conversely, if the desire is overturned by community, the social will be maintained. The goal of the fantasy which has the relationships of society is revealed through the various expressions of existence. They are divided into the subject and the other show the inner side of the main character. The subject shows the inner side of the main character by the transformation, alter ego and the other exists. The other shows the desire of the subject by the transformation, alter ego and the strangers. Disney animation studios select the target audience and the message in relation with the society. They choose the original like the fairy tale, myth and change them to satisfy the middle class. The characters of Disney animation says that messages by the expression of fantasy. The subject go through the transformation by twice. The first transformation is antisocial and the second transformation is social. The second shows a complete transformation. The other characters personified show the many kinds of the main character. The other as the alter ego of the main character represents the desire of the subject. They are described as an object of fear and exclusion. They expresses as a dark and menacing looks and hinders the complete transformation of the subject. But they were overthrown by the subject at the end of the story and it strengthen the social ideology. As a result, Disney highlights the value and the moral message of the society by using the representation of fantasy.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.6
• /
• pp.455-464
• /
• 2018

This study investigated whether the whitening effects of fullerene serum occurred through improvement of melanin production and skin tone and evaluated its potential for use as a cosmetic material for adult women aged 30-55 with dull skin color and pigmentation. Subjects were divided into a control group with fullerene-less serum (B) and an experimental group with serum containing 0.1% fullerene (A) and evaluated over 8 weeks. Visual assessment and measurement of skin conditions (melanin index, skin brightness) were conducted before use of and 4 and 8 weeks after use of the product, and a questionnaire survey and safety assessment were conducted at the end of the experiment. The results revealed that the highest decrease was 0.37% in the experimental group (A) after eight weeks of use of the product, but that this difference was not statistically significant. Moreover, no significant differences were observed upon inter-group comparison. Evaluation of changes in the melanin index and inter-group comparison revealed a significant decline in the experimental group (A) at all time points. Additionally, the $L^*$ value (skin brightness) of the experimental group (A) showed a significant increase (improvement) after eight weeks, while in inter-group comparison revealed a significant increase (improvement) at all time points (4 and 8 weeks later) (p<0.05). Based on a questionnaire survey of efficacy, positive response rates such as improvement of the target parts on week 8 were high in terms of 'improvement of skin color.' Moreover, the safety assessment revealed no adverse reactions. In conclusion, serum containing 0.1% fullerene revealed improved melanin production and skin tone (brightness) and confirmed its potential for use as a cosmetic material with brightening efficacy. Taken together, the results of this study indicate systematic investigation of the effective application and cosmetics formulations of this product are warranted.

• Progress in Medical Physics
• /
• v.18 no.2
• /
• pp.98-106
• /
• 2007

In June 2005, Yonsei University Medical Center, Severance Hospital upgraded a full-PACS system by adding twenty (5 mega pixels) Totoku ME511L flat panel LCD display devices for diagnostic interpretation purposes. Here we report upon the quantitative (or visual) acceptance testing of the twenty Totoku ME511L display devices for reflection, luminance response, luminance spatial dependency, resolution, noise, veiling glare, and display chromaticity based on AAPM TG 18 report. The tools used in the tests included a telescopic photometer, which was used as a colorimeter, illuminance meter, light sources for reflection assessment, light-blocking devices, and digital TG18 test patterns. For selected 8 flat panel displays, mean diffuse reflection coefficient ($R_d$) was $0.019{\pm}0.02sr^{-1}$. In the luminance response test, luminance ratio (LR), maximum luminance difference ($L_{max}$), and deviation of contrast response were $550{\pm}100,\;2.0{\pm}1.9%\;and\;5.8{\pm}1.8%$, respectively. In the luminance uniformity test, maximum luminance deviation was $14.3{\pm}5.5%$ for the 10% luminance of the TG18-UNL10 test pattern. In the resolution test with luminance measurement method, percent luminance (${\Dalta}L$) at the center was $0.94{\pm}0.64%$. In all cases of noise testing, rectangular target In every square in the three quadrants was visible and all 15 targets except the smallest one in the every corner pattern and the center pattern. The glare ratio (GR) was $12,346{\pm}1,995$. The color uniformity, (u',v'), was $0.0025{\pm}0.0008$. Also, the research results of qualify control guideline of primary disply devices suitable for domestic circumstance are presented All test results are in-line with the criteria recommended by AAPM TG18 report and are thus fully acceptable for diagnostic image interpretation. As a result, the acceptance testing schedule described provides not only an acceptance standard but also guidelines for quality control, optimized viewing conditions, and a means for determining the upgrading time of LCD display devices for diagnostic interpretation.

• Tuberculosis and Respiratory Diseases
• /
• v.43 no.5
• /
• pp.720-727
• /
• 1996

Objectives: $^{201}TI$ - chloride, $^{99m}Tc$ - MIBI, $^{99m}Tc$(V) - DMSA SPECT has been used in distinguishing lung cancer from benign lesion. To compare the diagnostic efficacy of SPECT with these tumor - seeking agents, we perfonned three consecutive SPECT using $^{201}TI$, $^{99m}Tc$ - MIBI, $^{99m}Tc$(V) - DMSA in same subjects with a solitary pulmonary lesion. Methods: SPECT was carried out at 10min and 3hr for $^{201}TI$ after injection of 20mCi, and 2hr for $^{99m}Tc$ - MIBI and $^{99m}Tc$(V) - DMSA after injection of 20mCi, respectively, in 37 patients with a solitary pulmonary lesion(27 lung cancer and 10 benign diseases). In patients showing visual uptake on lesion site, we obtained the lesion - to - background(target lesion/contralateral normal lung) uptake ratio from transverse slice for each radionuclide and also calculated the retention index for $^{201}TI$. Results: The diagnostic sensitivity of $^{201}TI$, $^{99m}Tc$ - MIBI and $^{99m}Tc$(V) - DMSA SPECT to lung cancer was 100%, 96% and 73%, and the specificity was 40%, 70% and 70%, respectively. The low specificities for these agents were mainly due to high positive uptake in patients with active pulmonary tuberculosis. There were no significant differences in uptake ratios and retention index between malignant and benign lesions, and among the histologic types of lung cancer. Conclusion : $^{201}TI$ and $^{99m}Tc$ - MIBI showed higher sensitivity than $^{99m}Tc$(V) - DMSA for detecting lung cancer, but was of limited usefulness in distinguishing lung cancer from benign lesion due to low specificity, especially in area with a high prevalence of active pulmonary tuberculosis.

• Progress in Medical Physics
• /
• v.14 no.2
• /
• pp.90-98
• /
• 2003

Purpose : A new virtual simulation technique for craniospinal irradiation (CSI) that uses a CT-simulator was developed to improve the accuracy of field and shielding placement as well as patient positioning. Materials and Methods : A CT simulator (CT-SIM) and a 3-D conformal radiation treatment planning system (3D-CRT) were used to develop CSI. The head and neck were immobilized with a thermoplastic mask while the rest of the body was immobilized with a Vac-Loc. A volumetric image was then obtained with the CT simulator. In order to improve the reproducibility of the setup, datum lines and points were marked on the head and body. Virtual fluoroscopy was performed with the removal of visual obstacles, such as the treatment table or immobilization devices. After virtual simulation, the treatment isocenters of each field were marked on the body and on the immobilization devices at the conventional simulation room. Each treatment fields was confirmed by comparing the fluoroscopy images with the digitally reconstructed radiography (DRR) and digitally composited radiography (DCR) images from virtual simulation. Port verification films from the first treatment were also compared with the DRR/DCR images for geometric verification. Results : We successfully performed virtual simulations on 11 CSI patients by CT-SIM. It took less than 20 minutes to affix the immobilization devices and to obtain the volumetric images of the entire body. In the absence of the patient, virtual simulation of all fields took 20 min. The DRRs were in agreement with simulation films to within 5 mm. This not only reducee inconveniences to the patients, but also eliminated position-shift variables attendant during the long conventional simulation process. In addition, by obtaining CT volumetric image, critical organs, such as the eyes and the spinal cord, were better defined, and the accuracy of the port designs and shielding was improved. Differences between the DRRs and the portal films were less than 3 m in the vertebral contour. Conclusion : Our analysis showed that CT simulation of craniospinal fields was accurate. In addition, CT simulation reduced the duration of the patient's immobility. During the planning process. This technique can improve accuracy in field placement and shielding by using three-dimensional CT-aided localization of critical and target structures. Overall, it has improved staff efficiency and resource utilization by standard protocol for craniospinal irradiation.

• Journal of Intelligence and Information Systems
• /
• v.21 no.1
• /
• pp.99-118
• /
• 2015

Object-oriented programming languages have been widely selected for developing modern information systems. The use of concepts relating to object-oriented (OO, in short) programming has reduced efforts of reusing pre-existing codes, and the OO concepts have been proved to be a useful in interpreting system requirements. In line with this, we have witnessed that a modern conceptual modeling approach supports features of object-oriented programming. Unified Modeling Language or UML becomes one of de-facto standards for information system designers since the language provides a set of visual diagrams, comprehensive frameworks and flexible expressions. In a modeling process, UML users need to consider relationships between classes. Based on an explicit and clear representation of classes, the conceptual model from UML garners necessarily attributes and methods for guiding software engineers. Especially, identifying an association between a class of part and a class of whole is included in the standard grammar of UML. The representation of part-whole relationship is natural in a real world domain since many physical objects are perceived as part-whole relationship. In addition, even abstract concepts such as roles are easily identified by part-whole perception. It seems that a representation of part-whole in UML is reasonable and useful. However, it should be admitted that the use of UML is limited due to the lack of practical guidelines on how to identify a part-whole relationship and how to classify it into an aggregate- or a composite-association. Research efforts on developing the procedure knowledge is meaningful and timely in that misleading perception to part-whole relationship is hard to be filtered out in an initial conceptual modeling thus resulting in deterioration of system usability. The current method on identifying and classifying part-whole relationships is mainly counting on linguistic expression. This simple approach is rooted in the idea that a phrase of representing has-a constructs a par-whole perception between objects. If the relationship is strong, the association is classified as a composite association of part-whole relationship. In other cases, the relationship is an aggregate association. Admittedly, linguistic expressions contain clues for part-whole relationships; therefore, the approach is reasonable and cost-effective in general. Nevertheless, it does not cover concerns on accuracy and theoretical legitimacy. Research efforts on developing guidelines for part-whole identification and classification has not been accumulated sufficient achievements to solve this issue. The purpose of this study is to provide step-by-step guidelines for identifying and classifying part-whole relationships in the context of UML use. Based on the theoretical work on Meta-model Formalization, self-check forms that help conceptual modelers work on part-whole classes are developed. To evaluate the performance of suggested idea, an experiment approach was adopted. The findings show that UML users obtain better results with the guidelines based on Meta-model Formalization compared to a natural language classification scheme conventionally recommended by UML theorists. This study contributed to the stream of research effort about part-whole relationships by extending applicability of Meta-model Formalization. Compared to traditional approaches that target to establish criterion for evaluating a result of conceptual modeling, this study expands the scope to a process of modeling. Traditional theories on evaluation of part-whole relationship in the context of conceptual modeling aim to rule out incomplete or wrong representations. It is posed that qualification is still important; but, the lack of consideration on providing a practical alternative may reduce appropriateness of posterior inspection for modelers who want to reduce errors or misperceptions about part-whole identification and classification. The findings of this study can be further developed by introducing more comprehensive variables and real-world settings. In addition, it is highly recommended to replicate and extend the suggested idea of utilizing Meta-model formalization by creating different alternative forms of guidelines including plugins for integrated development environments.