• Journal of Convergence for Information Technology
• /
• v.10 no.8
• /
• pp.23-34
• /
• 2020

Recently, the artificial intelligence deep learning field has been hard to commercialize due to the high computing power and the price problem of computing resources. In this paper, we apply a double pruning techniques to evaluate the performance of the in-depth neural network and various datasets. Double pruning combines basic Network-slimming and Parameter-prunning. Our proposed technique has the advantage of reducing the parameters that are not important to the existing learning and improving the speed without compromising the learning accuracy. After training various datasets, the pruning ratio was increased to reduce the size of the model.We confirmed that MobileNet-V3 showed the highest performance as a result of NetScore performance analysis. We confirmed that the performance after pruning was the highest in MobileNet-V3 consisting of depthwise seperable convolution neural networks in the Cifar 10 dataset, and VGGNet and ResNet in traditional convolutional neural networks also increased significantly.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.13 no.5
• /
• pp.313-322
• /
• 2020

A large volume of data will improve the robustness of deep learning models and avoid overfitting problems. In automatic tongue segmentation, the availability of annotated tongue images is often limited because of the difficulty of collecting and labeling the tongue image datasets in reality. Data augmentation can expand the training dataset and increase the diversity of training data by using label-preserving transformations without collecting new data. In this paper, augmented tongue image datasets were developed using seven augmentation techniques such as image cropping, rotation, flipping, color transformations. Performance of the data augmentation techniques were studied using state-of-the-art transfer learning models, for instance, InceptionV3, EfficientNet, ResNet, DenseNet and etc. Our results show that geometric transformations can lead to more performance gains than color transformations and the segmentation accuracy can be increased by 5% to 20% compared with no augmentation. Furthermore, a random linear combination of geometric and color transformations augmentation dataset gives the superior segmentation performance than all other datasets and results in a better accuracy of 94.98% with InceptionV3 models.

• Geomechanics and Engineering
• /
• v.37 no.5
• /
• pp.475-498
• /
• 2024

The prediction of the susceptibility of soil to liquefaction using a limited set of parameters, particularly when dealing with highly unbalanced databases is a challenging problem. The current study focuses on different ensemble learning classification algorithms using highly unbalanced databases of results from in-situ tests; standard penetration test (SPT), shear wave velocity (V_s) test, and cone penetration test (CPT). The input parameters for these datasets consist of earthquake intensity parameters, strong ground motion parameters, and in-situ soil testing parameters. liquefaction index serving as the binary output parameter. After a rigorous comparison with existing literature, extreme gradient boosting (XGBoost), bagging, and random forest (RF) emerge as the most efficient models for liquefaction instance classification across different datasets. Notably, for SPT and V_s-based models, XGBoost exhibits superior performance, followed by Light gradient boosting machine (LightGBM) and Bagging, while for CPT-based models, Bagging ranks highest, followed by Gradient boosting and random forest, with CPT-based models demonstrating lower G_mean(error), rendering them preferable for soil liquefaction susceptibility prediction. Key parameters influencing model performance include internal friction angle of soil (ϕ) and percentage of fines less than 75 µ (F₇₅) for SPT and V_s data and normalized average cone tip resistance (q_c) and peak horizontal ground acceleration (a_max) for CPT data. It was also observed that the addition of V_s measurement to SPT data increased the efficiency of the prediction in comparison to only SPT data. Furthermore, to enhance usability, a graphical user interface (GUI) for seamless classification operations based on provided input parameters was proposed.

• Structural Engineering and Mechanics
• /
• v.91 no.5
• /
• pp.443-457
• /
• 2024

At present, the traditional concrete surface inspection methods based on artificial vision have the problems of high cost and insecurity, while the computer vision methods rely on artificial selection features in the case of sensitive environmental changes and difficult promotion. In order to solve these problems, this paper introduces deep learning technology in the field of computer vision to achieve automatic feature extraction of structural damage, with excellent detection speed and strong generalization ability. The main contents of this study are as follows: (1) A method based on DeepLabV3+ convolutional neural network model is proposed for surface detection of post-earthquake structural damage, including surface damage such as concrete cracks, spaling and exposed steel bars. The key semantic information is extracted by different backbone networks, and the data sets containing various surface damage are trained, tested and evaluated. The intersection ratios of 54.4%, 44.2%, and 89.9% in the test set demonstrate the network's capability to accurately identify different types of structural surface damages in pixel-level segmentation, highlighting its effectiveness in varied testing scenarios. (2) A semantic segmentation model based on DeepLabV3+ convolutional neural network is proposed for the detection and evaluation of post-earthquake structural components. Using a dataset that includes building structural components and their damage degrees for training, testing, and evaluation, semantic segmentation detection accuracies were recorded at 98.5% and 56.9%. To provide a comprehensive assessment that considers both false positives and false negatives, the Mean Intersection over Union (Mean IoU) was employed as the primary evaluation metric. This choice ensures that the network's performance in detecting and evaluating pixel-level damage in post-earthquake structural components is evaluated uniformly across all experiments. By incorporating deep learning technology, this study not only offers an innovative solution for accurately identifying post-earthquake damage in civil engineering structures but also contributes significantly to empirical research in automated detection and evaluation within the field of structural health monitoring.

• Journal of the Korean Society of Radiology
• /
• v.17 no.1
• /
• pp.37-46
• /
• 2023

This study aims to evaluate the usefulness of Deep Learning Image Reconstruction (TrueFidelity, TF), the image quality of existing Filtered Back Projection (FBP) and Adaptive Statistical Iterative Reconstruction-Veo (ASIR-V) were compared. Noise, CNR, and SSIM were measured by obtaining images with doses fixed at 17.29 mGy and altered to 10.37 mGy, 12.10 mGy, 13.83 mGy, and 15.56 mGy in reconstruction techniques of FBP, ASIR-V 50%, and TF-H. TF-H has superior image quality compared to FBP and ASIR-V when the reconstruction technique change is given at 17.29 mGy. When dose changes were made, Noise, CNR, and SSIM were significantly different when comparing 10.37 mGy TF-H and FBP (p<0.05), and no significant difference when comparing 10.37 mGy TF-H and ASIR-V 50% (p>0.05). TF-H has a dose-reduction effect of 30%, as the highest dose of 15.56 mGy ASIR-V has the same image quality as the lowest dose of 10.37 mGy TF-H. Thus, Deep Learning Reconstruction techniques (TF) were able to reduce dose compared to Iterative Reconstruction techniques (ASIR-V) and Filtered Back Projection (FBP). Therefore, it is considered to reduce the exposure dose of patients.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.6
• /
• pp.1462-1477
• /
• 2024

With the evolving complexity of connected vehicle features, the volume and diversity of data generated during driving continue to escalate. Enabling data sharing among interconnected vehicles holds promise for improving users' driving experiences and alleviating traffic congestion. Yet, the unintentional disclosure of users' private information through data sharing poses a risk, potentially compromising the interests of vehicle users and, in certain cases, endangering driving safety. Federated learning (FL) is a newly emerged distributed machine learning paradigm, which is expected to play a prominent role for privacy-preserving learning in autonomous vehicles. While FL holds significant potential to enhance the architecture of the Internet of Vehicles (IoV), the dynamic mobility of vehicles poses a considerable challenge to integrating FL with vehicular networks. In this paper, a novel clustered FL framework is proposed which is efficient for reducing communication and protecting data privacy. By assessing the similarity among feature vectors, vehicles are categorized into distinct clusters. An optimal vehicle is elected as the cluster head, which enhances the efficiency of personalized data processing and model training while reducing communication overhead. Simultaneously, the Local Differential Privacy (LDP) mechanism is incorporated during local training to safeguard vehicle privacy. The simulation results obtained from the 20newsgroups dataset and the MNIST dataset validate the effectiveness of the proposed scheme, indicating that the proposed scheme can ensure data privacy effectively while reducing communication overhead.

• Asia-Pacific Journal of Business
• /
• v.3 no.1
• /
• pp.1-9
• /
• 2012

The Index of Learning Styles (ILS) instrument based on the Felder-Silverman Learning Style Model was used to determine distribution of learning styles of 125 South Korean business students enrolled in a South Korean institution of higher education. Results show that greater proportion of South Korean business students surveyed in this study prefer sensing over intuitive, visual over verbal, reflective over active, and global over sequential learning styles. The majority of business students have a balanced learning style in all four dimensions of the Felder-Silverman model. Among the students that do not have a balanced learning style, students with sensing, visual, reflective, and global learning styles dominate. Gender difference in learning style preference was not statistically significant for any of the four dimensions.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.12
• /
• pp.9-16
• /
• 2019

The goal of this study is to propose an efficient model for recognizing and classifying tree images to measure the accuracy that can be applied to smart devices during class. From the 2009 revised textbook to the 2015 revised textbook, the learning objective to the fourth-grade science textbook of elementary schools was added to the plant recognition utilizing smart devices. In this study, we compared the recognition rates of trees before and after retraining using a pre-trained inception V3 model, which is the support of the Google Inception V3. In terms of tree recognition, it can distinguish several features, including shapes, bark, leaves, flowers, and fruits that may lead to the recognition rate. Furthermore, if all the leaves of trees may fall during winter, it may challenge to identify the type of tree, as only the bark of the tree will remain some leaves. Therefore, the effective tree classification model is presented through the combination of the images by tree type and the method of combining the model for the accuracy of each tree type. I hope that this model will apply to smart devices used in educational settings.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.3
• /
• pp.570-590
• /
• 2024

Breast cancer ranks among the most prevalent forms of malignancy and foremost cause of death by cancer worldwide. It is not preventable. Early and precise detection is the only remedy for lowering the rate of mortality and improving the probability of survival for victims. In contrast to present procedures, thermography aids in the early diagnosis of cancer and thereby saves lives. But the accuracy experiences detrimental impact by low sensitivity for small and deep tumours and the subjectivity by physicians in interpreting the images. Employing deep learning approaches for cancer detection can enhance the efficacy. This study explored the utilization of thermography in early identification of breast cancer with the use of a publicly released dataset known as the DMR-IR dataset. For this purpose, we employed a novel approach that entails the utilization of a pre-trained MobileNetV2 model and fine tuning it through transfer learning techniques. We created three models using MobileNetV2: one was a baseline transfer learning model with weights trained from ImageNet dataset, the second was a fine-tuned model with an adaptive learning rate, and the third utilized early stopping with callbacks during fine-tuning. The results showed that the proposed methods achieved average accuracy rates of 85.15%, 95.19%, and 98.69%, respectively, with various performance indicators such as precision, sensitivity and specificity also being investigated.

• Journal of IKEEE
• /
• v.27 no.4
• /
• pp.379-385
• /
• 2023

Distributed congestion control (DCC) is a technology that mitigates channel congestion and improves communication performance in high-density vehicular networks. Traditional DCC techniques operate to reduce channel congestion without considering quality of service (QoS) requirements. Such design of DCC algorithms can lead to excessive DCC actions, potentially degrading other aspects of QoS. To address this issue, we propose a deep reinforcement learning-based QoS-adaptive DCC algorithm. The simulation was conducted using a quasi-real environment simulator, generating dynamic vehicular densities for evaluation. The simulation results indicate that our proposed DCC algorithm achieves results closer to the targeted QoS compared to existing DCC algorithms.