• Journal of Internet Computing and Services
• /
• v.19 no.6
• /
• pp.91-99
• /
• 2018

Sketch-based 3D object retrieval is a convenient way to search for various 3D data using human-drawn sketches as query. In this paper, we propose a new method of using Sketch CNN, Wasserstein CNN and Wasserstein center loss for sketch-based 3D object search. Specifically, Wasserstein center loss is a method of learning the center of each object category and reducing the Wasserstein distance between center and features of the same category. To do this, the proposed 3D object retrieval is performed as follows. Firstly, Wasserstein CNN extracts 2D images taken from various directions of 3D object using CNN, and extracts features of 3D data by computing the Wasserstein barycenters of features of each image. Secondly, the features of the sketch are extracted using a separate Sketch CNN. Finally, we learn the features of the extracted 3D object and the features of the sketch using the proposed Wasserstein center loss. In order to demonstrate the superiority of the proposed method, we evaluated two sets of benchmark data sets, SHREC 13 and SHREC 14, and the proposed method shows better performance in all conventional metrics compared to the state of the art methods.

• Journal of Internet Computing and Services
• /
• v.22 no.1
• /
• pp.23-30
• /
• 2021

This paper proposes a noise-tolerant image classification system using multiple autoencoders. The development of deep learning technology has dramatically improved the performance of image classifiers. However, if the images are contaminated by noise, the performance degrades rapidly. Noise added to the image is inevitably generated in the process of obtaining and transmitting the image. Therefore, in order to use the classifier in a real environment, we have to deal with the noise. On the other hand, the autoencoder is an artificial neural network model that is trained to have similar input and output values. If the input data is similar to the training data, the error between the input data and output data of the autoencoder will be small. However, if the input data is not similar to the training data, the error will be large. The proposed system uses the relationship between the input data and the output data of the autoencoder, and it has two phases to classify the images. In the first phase, the classes with the highest likelihood of classification are selected and subject to the procedure again in the second phase. For the performance analysis of the proposed system, classification accuracy was tested on a Gaussian noise-contaminated MNIST dataset. As a result of the experiment, it was confirmed that the proposed system in the noisy environment has higher accuracy than the CNN-based classification technique.

• Geophysics and Geophysical Exploration
• /
• v.24 no.3
• /
• pp.89-97
• /
• 2021

Fault detection in seismic data is well suited to the application of machine learning algorithms. Accordingly, various machine learning techniques are being developed. In recent studies, machine learning models, which utilize synthetic data, are the particular focus when training with deep learning. The use of synthetic training data has many advantages; Securing massive data for training becomes easy and generating exact fault labels is possible with the help of synthetic training data. To interpret real data with the model trained by synthetic data, the synthetic data used for training should be geologically realistic. In this study, we introduce a method to generate realistic synthetic seismic data. Initially, reflectivity models are generated to include realistic fault structures, and then, a one-way wave equation is applied to efficiently generate seismic stack sections. Next, a migration algorithm is used to remove diffraction artifacts and random noise is added to mimic actual field data. A convolutional neural network model based on the U-Net structure is used to verify the generated synthetic data set. From the results of the experiment, we confirm that realistic synthetic data effectively creates a deep learning model that can be applied to field data.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.139-147
• /
• 2021

This paper introduces and implements a Sound Event Detection (SED) system based on weakly-supervised learning where only part of the data is labeled, and analyzes the effect of parameters. The SED system estimates the classes and onset/offset times of events in the acoustic signal. In order to train the model, all information on the event class and onset/offset times must be provided. Unfortunately, the onset/offset times are hard to be labeled exactly. Therefore, in the weakly-supervised task, the SED model is trained by "strongly labeled data" including the event class and activations, "weakly labeled data" including the event class, and "unlabeled data" without any label. Recently, the SED systems using the mean-teacher model are widely used for the task with several parameters. These parameters should be chosen carefully because they may affect the performance. In this paper, performance analysis was performed on parameters, such as the feature, moving average parameter, weight of the consistency cost function, ramp-up length, and maximum learning rate, using the data of DCASE 2020 Task 4. Effects and the optimal values of the parameters were discussed.

• The Korean Journal of Franchise Management
• /
• v.10 no.2
• /
• pp.59-70
• /
• 2019

Purpose - Every company studies how to attract and retain new customers to increase competitiveness and profitability. Companies establish strategies to attract customers, secure competitive advantage and generate revenue. Businesses are looking for newer and better ways to differentiate themselves in the marketplace. One of the requirements for service differentiation is to make it a prerequisite for an engaging customer experience. Customer experience can be attained through service experience. Satisfaction determine whether to reuse the food service franchise. The purpose of this study is to investigate the effect of customer experience on the satisfaction and revisit intention of food service franchise. In this study, customer experience consists of three attributes such as service environment, food quality, and price fairness. Also, this study is to identify the importance of three service experience attributes of customer satisfaction and revisit intention using ANN (artificial neural network) analysis. Research design, data, methodology - The survey was conducted on customers who have visited franchise restaurants in one month in order to examine how service environment, food quality, and price fairness have been influenced customer satisfaction and revisit intention through online survey company (SM culture & contents). A total of 300 representative surveys were collected. Of those collected surveys, 26 were not used due to missing information, resulting in 274 as the final sample size. The sample size was more than 10 times more than the number of variables used in the structural model analysis. Results - The findings of this study are as follows: Service environment and price fairness have a significant effect on satisfaction. However, food quality did not have a significant effect on satisfaction. Finally, it was found that satisfaction had a significant effect on revisit intention. Meanwhile, according to the results of ANN analysis, satisfaction as a dependent variable was found to be the most important in male price fairness and service environment in female. Also, when the revisit intention is used as a dependent variable, both male and female price fairness are important. Also, when the intention to revisit is used as a dependent variable, both male and female price processes are important. Conclusions - First, a restaurant franchise enterprise needs to manage customer service experience. Customers should strive to eat and enjoy at a dining franchise store. Second, it is necessary to design a food service franchise shop as a customer-oriented service environment. Franchise companies need to improve the environment so that customers can use the store conveniently. Third, the restaurant franchise menu price needs to be cheaper than the alternative menu. The restaurant franchise menu needs to be constructed with a popular menu that can be used continuously by the customer, so that it can be set at a reasonable price.

• Journal of the Korea Convergence Society
• /
• v.12 no.3
• /
• pp.187-196
• /
• 2021

The paper deals with comparative study for characteristics of approximation of design space according to various approximate models and sensitivity analysis using orthogonal array experiments in structure design of active type DSF which was developed for float-over installation of offshore plant. This study aims to propose the orthogonal array experiments based design methodology which is able to efficiently explore an optimum design case and to generate the accurate approximate model. Thickness sizes of main structure member were applied to the design factors, and output responses were considered structure weight and strength performances. Quantitative effects on the output responses for each design factor were evaluated using the orthogonal array experiment. Best design case was also identified to improve the structure design with weight minimization. From the orthogonal array experiment results, various approximate models such as response surface model, Kriging model, Chebyshev orthogonal polynomial model, and radial basis function based neural network model were generated. The experiment results from orthogonal array method were validated by the approximate modeling results. It was found that the radial basis function based neural network model among the approximate models was able to approximate the design space of the active type DSF with the highest accuracy.

• Journal of IKEEE
• /
• v.25 no.2
• /
• pp.243-251
• /
• 2021

In this paper, we propose a measurement technique of indoor location based on markerless applicable to AR. The proposed technique has the following originality. The first is to extract feature points and use them to generate local patches to enable faster computation by learning and using only local patches that are more useful than the surroundings without learning the entire image. Second, learning is performed through deep learning using the convolution neural network structure to improve accuracy by reducing the error rate. Third, unlike the existing feature point matching technique, it enables indoor location measurement including left and right movement. Fourth, since the indoor location is newly measured every frame, errors occurring in the front side during movement are prevented from accumulating. Therefore, it has the advantage that the error between the final arrival point and the predicted indoor location does not increase even if the moving distance increases. As a result of the experiment conducted to evaluate the time required and accuracy of the measurement technique of indoor location based on markerless applicable to AR proposed in this paper, the difference between the actual indoor location and the measured indoor location is an average of 12.8cm and a maximum of 21.2cm. As measured, the indoor location measurement accuracy was better than that of the existing IEEE paper. In addition, it was determined that it was possible to measure the user's indoor location in real time by displaying the measured result at 20 frames per second.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.6
• /
• pp.573-581
• /
• 2020

Using CCTV images and weather parameters, a method for estimating PM (Particulate Matter) index was proposed, and an experiment was conducted. For CCTV images, we proposed a method of estimating the PM index by applying a deep learning technique based on a CNN (Convolutional Neural Network) with ROI(Region Of Interest) image including a specific spot and an full area image. In addition, after combining the predicted result values by deep learning with the two weather parameters of humidity and wind speed, a post-processing experiment was also conducted to calculate the modified PM index using the learned regression model. As a result of the experiment, the estimated value of the PM index from the CCTV image was R2(R-Squared) 0.58~0.89, and the result of learning the ROI image and the full area image with the measuring device was the best. The result of post-processing using weather parameters did not always show improvement in accuracy in all cases in the experimental area.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.6
• /
• pp.1291-1300
• /
• 2020

Recently, a deep learning-based non-profiling side-channel analysis was proposed. The deep learning-based non-profiling analysis is a technique that trains a neural network model for all guessed keys and then finds the correct secret key through the difference in the training metrics. As the performance of non-profiling analysis varies greatly depending on the neural network training model design, a correct model design criterion is required. This paper describes the two types of loss functions and eight labeling methods used in the training model design. It predicts the analysis performance of each labeling method in terms of non-profiling analysis and power consumption model. Considering the characteristics of non-profiling analysis and the HW (Hamming Weight) power consumption model is assumed, we predict that the learning model applying the HW label without One-hot encoding and the Correlation Optimization (CO) loss will have the best analysis performance. And we performed actual analysis on three data sets that are Subbytes operation part of AES-128 1 round. We verified our prediction by non-profiling analyzing two data sets with a total 16 of MLP-based model, which we describe.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.12
• /
• pp.320-330
• /
• 2020

Fire is difficult to achieve good performance in image detection using deep learning because of its high irregularity. In particular, there is little data on fire detection in underground utility facilities, which have poor light conditions and many objects similar to fire. These make fire detection challenging and cause low performance of deep learning models. Therefore, this study proposed a fire detection model using deep learning and estimated the performance of the model. The proposed model was designed using a combination of a basic convolutional neural network, Inception block of GoogleNet, and Skip connection of ResNet to optimize the deep learning model for fire detection under underground utility facilities. In addition, a training technique for the model was proposed. To examine the effectiveness of the method, the trained model was applied to fire images, which included fire and non-fire (which can be misunderstood as a fire) objects under the underground facilities or similar conditions, and results were analyzed. Metrics, such as precision and recall from deep learning models of other studies, were compared with those of the proposed model to estimate the model performance qualitatively. The results showed that the proposed model has high precision and recall for fire detection under low light intensity and both low erroneous and missing detection capabilities for things similar to fire.