• The Journal of Korea Robotics Society
• /
• v.14 no.1
• /
• pp.31-39
• /
• 2019

A robot usually adopts ANN (artificial neural network)-based object detection and instance segmentation algorithms to recognize objects but creating datasets for these algorithms requires high labeling costs because the dataset should be manually labeled. In order to lower the labeling cost, a new scheme is proposed that can automatically generate a training images and label them for specific objects. This scheme uses an instance segmentation algorithm trained to give the masks of unknown objects, so that they can be obtained in a simple environment. The RGB images of objects can be obtained by using these masks, and it is necessary to label the classes of objects through a human supervision. After obtaining object images, they are synthesized with various background images to create new images. Labeling the synthesized images is performed automatically using the masks and previously input object classes. In addition, human intervention is further reduced by using the robot arm to collect object images. The experiments show that the performance of instance segmentation trained through the proposed method is equivalent to that of the real dataset and that the time required to generate the dataset can be significantly reduced.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.18 no.6
• /
• pp.277-283
• /
• 2018

Some innovative researchers have had a dream of self-thinking intelligent computer. Alphago, at last, showed its possibility. With it, most computer engineers including even students can learn easily how to do it. As the interest to the deep learning has been growing, people's expectation is also naturally growing. In this research, we tried to enhance the game ability of a 7-poker system by applying machine learning techniques. In addition, we also tried to apply emotion analysis of a player to trace ones emotional changes. Methods and outcomes are to be explained in this paper.

• Journal of Biomedical Engineering Research
• /
• v.40 no.2
• /
• pp.48-54
• /
• 2019

This paper proposed a method of validation data augmentation for improving the grading accuracy of diabetic macular edema (DME) using deep learning. The data augmentation technique is basically applied in order to secure diversity of data by transforming one image to several images through random translation, rotation, scaling and reflection in preparation of input data of the deep neural network (DNN). In this paper, we apply this technique in the validation process of the trained DNN, and improve the grading accuracy by combining the classification results of the augmented images. To verify the effectiveness, 1,200 retinal images of Messidor dataset was divided into training and validation data at the ratio 7:3. By applying random augmentation to 359 validation data, $1.61{\pm}0.55%$ accuracy improvement was achieved in the case of six times augmentation (N=6). This simple method has shown that the accuracy can be improved in the N range from 2 to 6 with the correlation coefficient of 0.5667. Therefore, it is expected to help improve the diagnostic accuracy of DME with the grading information provided by the proposed DNN.

• Journal of Information Processing Systems
• /
• v.14 no.6
• /
• pp.1508-1520
• /
• 2018

In this paper, we propose an improved model to provide users with a better long-term prediction of waterworks operation data. The existing prediction models have been studied in various types of models such as multiple linear regression model while considering time, days and seasonal characteristics. But the existing model shows the rate of prediction for demand fluctuation and long-term prediction is insufficient. Particularly in the deep running model, the long-short-term memory (LSTM) model has been applied to predict data of water purification plant because its time series prediction is highly reliable. However, it is necessary to reflect the correlation among various related factors, and a supplementary model is needed to improve the long-term predictability. In this paper, convolutional neural network (CNN) model is introduced to select various input variables that have a necessary correlation and to improve long term prediction rate, thus increasing the prediction rate through the LSTM predictive value and the combined structure. In addition, a multiple linear regression model is applied to compile the predicted data of CNN and LSTM, which then confirms the data as the final predicted outcome.

• Journal of the Korean Geosynthetics Society
• /
• v.17 no.4
• /
• pp.251-265
• /
• 2018

This paper concerns the development of an optimized NATM secondary lining design system for a subsea tunnel. The subsea tunnel is normally laid down under the sea water and submarine ground which consists of soil or rock. The design system is the series of process which can predict lining member forces by ANN (artificial neural network system), analyze suitable section for the designated ground, construction and tunnel conditions. Finally, this lining design system aims to be connected for designing the subsea tunnel automatically. The lining member forces are predicted based on the ANN which was calculated by a FEM (finite element analysis) and it helps designers determine its lining dimension easily without any further FEM calculations.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.18 no.3
• /
• pp.74-81
• /
• 2019

Herein, we performed a comparative study on approximate multi-objective design optimization, to realize a structural design to improve the weight and vibration performances of the knuckle - a car suspension component - considering various load conditions and vibration characteristics. In the approximate multi-objective optimization process, a regression meta-model was generated using the response surfaces method (RSM), while Kriging and back-propagation neural network (BPN) methods were applied for interpolation meta-modeling. The Pareto solutions, multi-objective optimal solutions, were derived using the non-dominated sorting genetic algorithm (NSGA-II). In terms of the knuckle design considered in this study, the characteristics and influence of the meta-model on multi-objective optimization were reviewed through a comparison of the approximate optimization results with the meta-models and the actual optimization.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.14 no.4
• /
• pp.69-77
• /
• 2018

The malignant melanoma accounts for about 1 to 3% of the total malignant tumor in the West, especially in the US, it is a disease that causes more than 9,000 deaths each year. Generally, skin lesions are difficult to detect the features through photography. In this paper, we propose a computer-aided diagnosis algorithm based on deep learning for classification of malignant melanoma and benign skin tumor in RGB channel skin images. The proposed deep learning model configures the tumor lesion segmentation model and a classification model of malignant melanoma. First, U-Net was used to segment a skin lesion area in the dermoscopic image. We could implement algorithms to classify malignant melanoma and benign tumor using skin lesion image and results of expert's labeling in ResNet. The U-Net model obtained a dice similarity coefficient of 83.45% compared with results of expert's labeling. The classification accuracy of malignant melanoma obtained the 83.06%. As the result, it is expected that the proposed artificial intelligence algorithm will utilize as a computer-aided diagnosis algorithm and help to detect malignant melanoma at an early stage.

• Journal of Integrative Natural Science
• /
• v.11 no.4
• /
• pp.167-183
• /
• 2018

Electroencephalogram (EEG) recording provides a new way to support human-machine communication. It gives us an opportunity to analyze the neuro-dynamics of human cognition. Machine learning is a powerful for the EEG classification. In addition, machine learning can compensate for high variability of EEG when analyzing data in real time. However, the optimal EEG electrode location must be prioritized in order to extract the most relevant features from brain wave data. In this paper, we propose an intelligent system model for the extraction of EEG data by training the optimal electrode location of EEG in a specific problem. The proposed system is basically a fuzzy system and uses a neural network structurally. The fuzzy clustering method is used to determine the optimal number of fuzzy rules using the features extracted from the EEG data. The parameters and weight values found in the process of determining the number of rules determined here must be tuned for optimization in the learning process. Genetic algorithms are used to obtain optimized parameters. We present useful results by using optimal rule numbers and non - symmetric membership function using EEG data for four movements with the right arm through various experiments.

• Korean Journal of Remote Sensing
• /
• v.35 no.1
• /
• pp.93-115
• /
• 2019

The availability of high-resolution laser scanning data and advanced machine learning algorithms has enabled an accurate potential rockfall source identification. However, the presence of other mass movements, such as landslides within the same region of interest, poses additional challenges to this task. Thus, this research presents a method based on an integration of Gaussian mixture model (GMM) and ensemble artificial neural network (bagging ANN [BANN]) for automatic detection of potential rockfall sources at Kinta Valley area, Malaysia. The GMM was utilised to determine slope angle thresholds of various geomorphological units. Different algorithms(ANN, support vector machine [SVM] and k nearest neighbour [kNN]) were individually tested with various ensemble models (bagging, voting and boosting). Grid search method was adopted to optimise the hyperparameters of the investigated base models. The proposed model achieves excellent results with success and prediction accuracies at 95% and 94%, respectively. In addition, this technique has achieved excellent accuracies (ROC = 95%) over other methods used. Moreover, the proposed model has achieved the optimal prediction accuracies (92%) on the basis of testing data, thereby indicating that the model can be generalised and replicated in different regions, and the proposed method can be applied to various landslide studies.

• The Journal of the Convergence on Culture Technology
• /
• v.4 no.4
• /
• pp.261-266
• /
• 2018

This study aimed to provide an overview of artificial intelligence based music composition programs. The artificial intelligence-based composition program has shown remarkable growth as the development of deep neural network theory and the improvement of big data processing technology. Accordingly, artificial intelligence based composition programs for composing classical music and pop music have been proposed variously in academia and industry. But there are several limitations: devaluation in general populations, missing valuable materials, lack of relevant laws, technology-led industries exclusive to the arts, and so on. When effective measures are taken against these limitations, artificial intelligence based technology will play a significant role in fostering national competitiveness.