• The Journal of the Acoustical Society of Korea
• /
• v.43 no.1
• /
• pp.78-88
• /
• 2024

Ship-radiated noise received by passive sonar that can measure underwater noise can be identified and classified ship using Detection of Envelope Modulation on Noise (DEMON) analysis. However, in a low Signal-to-Noise Ratio (SNR) environment, it is difficult to analyze and identify the target frequency line containing ship information in the DEMONgram. In this paper, we conducted a study to extract target frequency lines using semantic segmentation among deep learning techniques for more accurate target identification in a low SNR environment. The semantic segmentation models U-Net, UNet++, and DeepLabv3+ were trained and evaluated using simulated DEMONgram data generated by changing SNR and fundamental frequency, and the DEMONgram prediction performance of DeepShip, a dataset of ship-radiated noise recordings on the strait of Georgia in Canada, was compared using the trained models. As a result of evaluating the trained model with the simulated DEMONgram, it was confirmed that U-Net had the highest performance and that it was possible to extract the target frequency line of the DEMONgram made by DeepShip to some extent.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.30 no.2
• /
• pp.147-156
• /
• 2024

The sustainable management and enhancement of marine resources are becoming increasingly important issues worldwide. This study was conducted in response to these challenges, focusing on the development and performance comparison of fish detection and classification models as part of a deep learning-based technique for assessing the effectiveness of marine resource enhancement projects initiated by the Korea Fisheries Resources Agency. The aim was to select the optimal model by training various sizes of YOLOv8-Seg models on a fish image dataset and comparing each performance metric. The dataset used for model construction consisted of 36,749 images and label files of 12 different species of fish, with data diversity enhanced through the application of augmentation techniques during training. When training and validating five different YOLOv8-Seg models under identical conditions, the medium-sized YOLOv8m-Seg model showed high learning efficiency and excellent detection and classification performance, with the shortest training time of 13 h and 12 min, an of 0.933, and an inference speed of 9.6 ms. Considering the balance between each performance metric, this was deemed the most efficient model for meeting real-time processing requirements. The use of such real-time fish detection and classification models could enable effective surveys of marine resource enhancement projects, suggesting the need for ongoing performance improvements and further research.

• Knowledge Management Research
• /
• v.24 no.1
• /
• pp.245-271
• /
• 2023

The change to the data economy requires a new analysis beyond ordinary research in the management field. Data matching refers to a technique or processing method that combines data sets collected from different samples with the same population. In this study, statistical matching was performed using random hotdeck and Mahalanobis distance functions using 2020 Survey of Korea Venture Firms and 2020 Korea Innovation Survey datas. Among the variables used for statistical matching simulation, the industry and the number of workers were set to be completely consistent, and region, business power, listed market, and sales were set as common variables. Simulation verification was confirmed by mean test and kernel density. As a result of the analysis, it was confirmed that statistical matching was appropriate because there was a difference in the average test, but a similar pattern was shown in the kernel density. This result attempted to expand the spectrum of the research method by experimenting with a data matching research methodology that has not been sufficiently attempted in the management field, and suggests implications in terms of data utilization and diversity.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.31 no.3
• /
• pp.229-238
• /
• 2013

Development of methodologies to generate the small scale map from the large scale map using map generalization has huge importance in management of the digital topographic map, such as producing and updating maps. In this study, the selection methodology of map generalization for the road network data in digital topographic map is investigated and evaluated. The existing maps with 1:5,000 and 1:25,000 scales are compared and the criteria for selection of the road network data, which are the number of objects and the relative importance of road network, are analyzed by using the T$\ddot{o}$pfer's radical law and Logit model. The selection model derived from the analysis result is applied to the test data, and the road network data of 1:18,000 and 1:72,000 scales from the digital topographic map of 1:5,000 scale are generated. The generalized results showed that the road objects with relatively high importance are selected appropriately according to the target scale levels after the qualitative and quantitative evaluations.

• Journal of KIISE
• /
• v.45 no.2
• /
• pp.113-125
• /
• 2018

Recently, with the advent of the Big Data era, we have gained the capability of acquiring vast amounts of knowledge from various fields. The collected knowledge is expressed by well-formed formula and in particular, OWL, a standard language of ontology, is a typical form of well-formed formula. The symbolic reasoning is actively being studied using large amounts of ontology data for extracting intrinsic information. However, most studies of this reasoning support the restricted rule expression based on Description Logic and they have limited applicability to the real world. Moreover, knowledge management for inaccurate information is required, since knowledge inferred from the wrong information will also generate more incorrect information based on the dependencies between the inference rules. Therefore, this paper suggests that the SWAT, knowledge management system should be combined with the SWRL (Semantic Web Rule Language) reasoning based on ATMS (Assumption-based Truth Maintenance System). Moreover, this system was constructed by combining with SWRL reasoning and ATMS for managing large ontology data based on the distributed In-memory framework. Based on this, the ATMS monitoring system allows users to easily detect and correct wrong knowledge. We used the LUBM (Lehigh University Benchmark) dataset for evaluating the suggested method which is managing the knowledge through the retraction of the wrong SWRL inference data on large data.

• Asia-Pacific Journal of Business Venturing and Entrepreneurship
• /
• v.16 no.3
• /
• pp.73-86
• /
• 2021

The primary purpose of this paper is to explore the advanced models that predict entrepreneurial innovativeness most accurately. For the first time in the field of entrepreneurship research, it presents a model that predicts entrepreneurial innovativeness based on machine learning corresponding to data scientific approaches. It uses 22,099 the Global Entrepreneurship Monitor (GEM) data from 62 countries to build predictive models. Based on the data set consisting of 27 explanatory variables, it builds predictive models that are traditional statistical methods such as multiple regression analysis and machine learning models such as regression tree, random forest, XG boost, and artificial neural networks. Then, it compares the performance of each model. It uses indicators such as root mean square error (RMSE), mean analysis error (MAE) and correlation to evaluate the performance of the model. The analysis of result is that all five machine learning models perform better than traditional methods, while the best predictive performance model was XG boost. In predicting it through XG boost, the variables with high contribution are entrepreneurial opportunities and cross-term variables of market expansion, which indicates that the type of entrepreneur who wants to acquire opportunities in new markets exhibits high innovativeness.

• The Journal of Bigdata
• /
• v.7 no.2
• /
• pp.31-41
• /
• 2022

Steel is one of the most fundamental components to mechanical industry. However, the quality of products are greatly impacted by the surface defects in the steel. Thus, researchers pay attention to the need for surface defects detector and the deep learning methods are the current trend of object detector. There are still limitations and rooms for improvements, for example, related works focus on developing the models but don't take into account real-time application with practical implication on industrial settings. In this paper, a real-time application of steel surface defects detection based on YOLOv4 is proposed. Firstly, as the aim of this work to deploying model on real-time application, we studied related works on this field, particularly focusing on one-stage detector and YOLO algorithm, which is one of the most famous algorithm for real-time object detectors. Secondly, using pre-trained Yolov4-Darknet platform models and transfer learning, we trained and test on the hot rolled steel defects open-source dataset NEU-DET. In our study, we applied our application with 4 types of typical defects of a steel surface, namely patches, pitted surface, inclusion and scratches. Thirdly, we evaluated YOLOv4 trained model real-time performance to deploying our system with accuracy of 87.1 % mAP@0.5 and over 60 fps with GPU processing.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• Journal of the Korean Society of Radiology
• /
• v.17 no.3
• /
• pp.423-431
• /
• 2023

The purpose of this study is to analyze the factors affecting the classification of the severity of contrast media side effects based on the patient's body information using artificial intelligence techniques to be used as basic data to reduce the degree of contrast medium side effects. The data used in this study were 606 examiners who had no contrast medium side effects in the past history survey among 1,235 cases of contrast medium side effects among 58,000 CT scans performed at a general hospital in Seoul. The total data is 606, of which 70% was used as a training set and the remaining 30% was used as a test set for validation. Age, BMI(Body Mass Index), GFR(Glomerular Filtration Rate), BUN(Blood Urea Nitrogen), GGT(Gamma Glutamyl Transgerase), AST(Aspartate Amino Transferase,), and ALT(Alanine Amiono Transferase) features were used as independent variables, and contrast media severity was used as a target variable. AUC(Area under curve), CA(Classification Accuracy), F1, Precision, and Recall were identified through AdaBoost, Tree, Neural network, SVM, and Random foest algorithm. AdaBoost and Random Forest show the highest evaluation index in the classification prediction algorithm. The largest factors in the predictions of all models were GFR, BMI, and GGT. It was found that the difference in the amount of contrast media injected according to renal filtration function and obesity, and the presence or absence of metabolic syndrome affected the severity of contrast medium side effects.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_3
• /
• pp.949-965
• /
• 2023

In this study, we propose a method to monitor the surface area of agricultural reservoirs in South Korea using Sentinel-1 synthetic aperture radar images and the deep learning model, Swin Transformer. Utilizing the Google Earth Engine platform, datasets from 2017 to 2021 were constructed for seven agricultural reservoirs, categorized into 700 K-ton, 900 K-ton, and 1.5 M-ton capacities. For four of the reservoirs, a total of 1,283 images were used for model training through shuffling and 5-fold cross-validation techniques. Upon evaluation, the Swin Transformer Large model, configured with a window size of 12, demonstrated superior semantic segmentation performance, showing an average accuracy of 99.54% and a mean intersection over union (mIoU) of 95.15% for all folds. When the best-performing model was applied to the datasets of the remaining three reservoirsfor validation, it achieved an accuracy of over 99% and mIoU of over 94% for all reservoirs. These results indicate that the Swin Transformer model can effectively monitor the surface area of agricultural reservoirs in South Korea.