• Journal of the Korea Convergence Society
• /
• v.9 no.12
• /
• pp.25-32
• /
• 2018

In the massive online fashion market, it is not easy for consumers to find the fashion style they want by keyword search for their preferred style. It can be resolved into consumer needs based fashion recommendation. Most of the existing online shopping sites have collected cumtomer's preference style using the online quastionnair. In this paper, we propose a simple but effective novel model that resolve the traditional method in fashion profiling for consumer's preference style and needs using implicit profiling method. In addition, we proposed a learning model that reflects the characteristics of the images itself through the deep learning-based intelligent preferred fashion model learned from the collected data. We show that the proposed model gave meaningful results through the qualitative evaluation.

• Journal of the Korea Computer Graphics Society
• /
• v.25 no.3
• /
• pp.123-131
• /
• 2019

In this study, we modified CT images of femoral head in consideration of anatomically meaningful structure, proposing the method to augment the training data of convolution Neural network for segmentation of femur mesh model. First, the femur mesh model is obtained from the CT image. Then divide the mesh model into meaningful parts by using cluster analysis on geometric characteristic of mesh surface. Finally, transform the segments by using an appropriate mesh deformation algorithm, then create new CT images by warping CT images accordingly. Deep learning models using the data enhancement methods of this study show better image division performance compared to data augmentation methods which have been commonly used, such as geometric conversion or color conversion.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.6
• /
• pp.682-689
• /
• 2019

In part-based image representation, the partial shapes of an object are represented as basis vectors, and an image is decomposed as a linear combination of basis vectors where the coefficients of those basis vectors represent the partial (or local) feature of an object. In this paper, a face recognition for occluded faces is proposed in which face images are represented using non-negative matrix factorization(NMF), one of part-based representation techniques, and recognized using an artificial neural network technique. Standard NMF, projected gradient NMF and orthogonal NMF were used in part-based representation of face images, and their performances were compared. Learning vector quantizer were used in the recognizer where Euclidean distance was used as the distance measure. Experimental results show that proposed recognition is more robust than the conventional face recognition for the occluded faces.

• Steel and Composite Structures
• /
• v.31 no.6
• /
• pp.629-640
• /
• 2019

The investigation of retaining wall structures behavior under dynamic loads is considered as one of important parts for designing such structures. Generally, the performance of these structures is under the influence of the environment conditions and their geometry. The aim of this research is to design retaining wall structures based on smart and optimal systems. The use of accuracy and speed to assess the structures under different conditions is one of the important parts sought by designers. Therefore, optimal and smart systems are able to have better addressing these problems. Using numerical and coding methods, this research investigates the retaining wall structure design under different dynamic conditions. More than 9500 models were constructed and considered for modelling design. These designs include height and thickness of the wall, soil density, rock density, soil friction angle, and peak ground acceleration (PGA) variables. Accordingly, a neural network system was developed to establish an appropriate relationship between data to obtain safety factor (SF) of retaining walls under different seismic conditions. Different parameters were analyzed and the effect of each parameter was assessed separately. According to these analyses, the structure optimization was performed to increase the SF values. The optimal and smart design showed that under different PGA conditions, the structure performance can be appropriately improved while utilization of the initial (or basic) parameters leads to the structure failure. Therefore, by increasing accuracy and speed, smart methods could improve the retaining structure performance in controlling the wall failure. The intelligent design process of this study can be applied to some other civil engineering applications such as slope stability.

• Journal of Digital Convergence
• /
• v.17 no.9
• /
• pp.71-78
• /
• 2019

This study aims to identify good borrowers within the context of P2P lending. P2P lending is a growing platform that allows individuals to lend and borrow money from each other. Inherent in any loans is credit risk of borrowers and needs to be considered before any lending. Specifically in the context of P2P lending, traditional models fall short and thus this study aimed to rectify this as well as explore the problem of class imbalances seen within credit risk data sets. This study implemented an over-sampling technique known as Synthetic Minority Over-sampling Technique (SMOTE). To test our approach, we implemented five benchmarking classifiers such as support vector machines, logistic regression, k-nearest neighbor, random forest, and deep neural network. The data sample used was retrieved from the publicly available LendingClub dataset. The proposed SMOTE revealed significantly improved results in comparison with the benchmarking classifiers. These results should help actors engaged within P2P lending to make better informed decisions when selecting potential borrowers eliminating the higher risks present in P2P lending.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.1
• /
• pp.27-36
• /
• 2019

In this paper, we propose a method to find the most similar answer to the user's response from the question-answer database in order to avoid generating a redundant question in retrieval-based automatic question generation system. As a question of the most similar answer to user's response may already be known to the user, the question should be removed from a set of question candidates. A similarity detector calculates a similarity between two answers by utilizing the same words, paraphrases, and sentential meanings. Paraphrases can be acquired by building a phrase table used in a statistical machine translation. A sentential meaning's similarity of two answers is calculated by an attention-based convolutional neural network. We evaluate the accuracy of the similarity detector on an evaluation set with 100 answers, and can get the 71% Mean Reciprocal Rank (MRR) score.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.1
• /
• pp.39-46
• /
• 2019

In this study, before system development for the elimination of foreign fish species, we propose an algorithm to classify fish species by training fish images with CNN. The raw data for CNN learning were directly captured images for each species, Dataset 1 increases the number of images to improve the classification of fish species and Dataset 2 realizes images close to natural environment are constructed and used as training and test data. The classification performance of four CNNs are over 99.97% for dataset 1 and 99.5% for dataset 2, in particular, we confirm that the learned CNN using Data Set 2 has satisfactory performance for fish images similar to the natural environment. And among four CNNs, AlexNet achieves satisfactory performance, and this has also the shortest execution time and training time, we confirm that it is the most suitable structure to develop the system for the elimination of foreign fish species.

• Journal of Broadcast Engineering
• /
• v.24 no.2
• /
• pp.217-226
• /
• 2019

In this paper, we propose a filtering method that can improve the results of inverse tone-mapping using guided image filter. Inverse tone-mapping techniques have been proposed that convert LDR images to HDR. Recently, many algorithms have been studied to convert single LDR images into HDR images using CNN. Among them, there exists an algorithm for restoring pixel information using CNN which learned to restore saturated region. The algorithm does not suppress the noise in the non-saturation region and cannot restore the detail in the saturated region. The proposed algorithm suppresses the noise in the non-saturated region and restores the detail of the saturated region using a WGIF in the input image, and then applies it to the CNN to improve the quality of the final image. The proposed algorithm shows a higher quantitative image quality index than the existing algorithms when the HDR quantitative image quality index was measured.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.11
• /
• pp.411-418
• /
• 2018

With the proliferation of artificial intelligence technology, it is becoming important to obtain, store, and utilize scientific data in research and science sectors. A number of methods for extracting meaningful objects such as graphs and tables from research articles have been proposed to eventually obtain scientific data. Existing extraction methods using heuristic approaches are hardly applicable to electronic documents having heterogeneous manuscript formats because they are designed to work properly for some targeted manuscripts. This paper proposes a prototype of an object extraction system which exploits a recent deep-learning technology so as to overcome the inflexibility of the heuristic approaches. We implemented our trained model, based on the Faster R-CNN algorithm, using the Google TensorFlow Object Detection API and also composed an annotated data set from 100 research articles for training and evaluation. Finally, a performance evaluation shows that the proposed system outperforms a comparator adopting heuristic approaches by 5.2%.

• Journal of Broadcast Engineering
• /
• v.23 no.6
• /
• pp.780-789
• /
• 2018

Nowadays, gun crimes occur very frequently not only in public places but in alleyways around the world. In particular, it is essential to detect a person armed by a pistol to prevent those crimes since small guns, such as pistols, are often used for those crimes. Because conventional works for armed person detection have treated an armed person as a single object in an input image, their accuracy is very low. The reason for the low accuracy comes from the fact that the gunman is treated as a single object although the pistol is a relatively much smaller object than the person. To solve this problem, we propose a novel algorithm called APDA(Armed Person Detection Algorithm). APDA detects the armed person using in a post-processing the positions of both wrists and the pistol achieved by the CNN-based human body feature detection model and the pistol detection model, respectively. We show that APDA can provide both 46.3% better recall and 14.04% better precision than SSD-MobileNet.