• Journal of the Korean Society of Radiology
• /
• v.18 no.4
• /
• pp.317-325
• /
• 2024

This study aimed to classify normal and nodule images in thyroid ultrasound images using GLCM and machine learning. The research was conducted on 600 patients who visited S Hospital in Busan and were diagnosed with thyroid nodules using thyroid ultrasound. In the thyroid ultrasound images, the ROI was set to a size of 50x50 pixels, and 21 parameters and 4 angles were used with GLCM to analyze the normal thyroid patterns and thyroid nodule patterns. The analyzed data was used to distinguish between normal and nodule diagnostic results using the SVM model and KNN model in MATLAB. As a result, the accuracy of the thyroid nodule classification rate was 94% for SVM model and 91% for the KNN model. Both models showed an accuracy of over 90%, indicating that the classification rate is excellent when using machine learning for the classification of normal thyroid and thyroid nodules. In the ROC curve, the ROC curve for the SVM model was generally higher compared to the KNN model, indicating that the SVM model has higher within-sample performance than the KNN model. Based on these results, the SVM model showed high accuracy in diagnosing thyroid nodules. This result can be used as basic data for future research as an auxiliary tool for medical diagnosis and is expected to contribute to the qualitative improvement of medical services through machine learning technology.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.17 no.4
• /
• pp.205-212
• /
• 2024

This study developed a deep learning model that distinguishes the front (with garnish) and the back (without garnish) surface of the dried semi-finished product (dried bukak) for screening operation before transfter the dried bukak to oil heater using robot's vacuum gripper. For deep learning model training and verification, RGB images for the front and back surfaces of 400 dry bukak that treated by data preproccessing were obtained. YOLO-v5 was used as a base structure of deep learning model. The area, surface information labeling, and data augmentation techniques were applied from the acquired image. Parameters including mAP, mIoU, accumulation, recall, decision, and F1-score were selected to evaluate the performance of the developed YOLO-v5 deep learning model-based surface detection model. The mAP and mIoU on the front surface were 0.98 and 0.96, respectively, and on the back surface, they were 1.00 and 0.95, respectively. The results of binary classification for the two front and back classes were average 98.5%, recall 98.3%, decision 98.6%, and F1-score 98.4%. As a result, the developed model can classify the surface information of the dried bukak using RGB images, and it can be used to develop a robot-automated system for the surface detection process of the dried bukak before deep frying.

• Journal of Platform Technology
• /
• v.11 no.5
• /
• pp.59-71
• /
• 2023

This paper proposes a method of estimating the illumination chromaticity of a scene in which an image is taken. Storring and Bianco proposed a method of estimating illuminant chromaticity using skin color. Storring et al. used skin color distribution characteristics and black body locus, but there is a problem that the link between the locus and CIE-xy data is reduced. Bianco et al. estimated the illuminant chromaticity by comparing the skin color distribution in standard lighting with the skin color distribution in the input image. This method is difficult to measure and secure as much skin color as possible in various illumination. The proposed method can estimate the illuminant chromaticity for any input image by analyzing the relationship between the skin color information and the illuminant chromaticity. The estimation method is divided into an analysis stage and a test stage, and the data set was classified into an analysis group and a test group and used. Skin chromaticity is calculated by obtaining skin color areas from all input images of the analysis group, respectively. A mapping is obtained by analyzing the correlation between the average set of skin chromaticity and the reference illuminant chromaticity set. The calculated mapping is applied to all input images of the analysis group to estimate the illuminant chromaticity, calculate the error with the reference illuminant chromaticity, and repeat the above process until there is no change in the error to obtain a stable mapping. The obtained mapping is applied to the test group images similar to the analysis stage to estimate the illuminant chromaticity. Since there is no independent data set containing skin area and illuminant reference information, the experimental data set was made using some of the images of the Intel TAU data set. Compared to Finlayson, a similar theory-based existing method, it showed performance improvement of more than 40%, Zhang 11%, and Kim 16%.

• The Journal of Bigdata
• /
• v.8 no.2
• /
• pp.73-82
• /
• 2023

With 78% of current fisheries workers being elderly, there's a pressing need to address labor shortages. Consequently, active research on smart aquaculture technologies, centered on object detection and tracking algorithms, is underway. These technologies allow for fish size analysis and behavior pattern forecasting, facilitating the development of real-time monitoring and automated systems. Our study utilized video data from cameras outside aquaculture facilities and implemented fish detection and tracking algorithms. We aimed to tackle high maintenance costs due to underwater conditions and camera corrosion from ammonia and pH levels. We evaluated the performance of a real-time system using YOLOv7 for fish detection and the SORT algorithm for movement tracking. YOLOv7 results demonstrated a trade-off between Recall and Precision, minimizing false detections from lighting, water currents, and shadows. Effective tracking was ascertained through re-identification. This research holds promise for enhancing smart aquaculture's operational efficiency and improving fishery facility management.

• Journal of the Korea Computer Graphics Society
• /
• v.30 no.3
• /
• pp.99-107
• /
• 2024

While the demand for extrapolating video content horizontally or vertically is increasing, even the most advanced techniques cannot successfully extrapolate all videos. Therefore, it is important to determine if a given video can be well extrapolated before attempting the actual extrapolation. This can help avoid wasting computing resources. This paper proposes a video classifier that can identify if a video is suitable for horizontal extrapolation. The classifier utilizes optical flow and an adaptive Gaussian blur network, which can be applied to flow-based video extrapolation methods. The labeling for training was rigorously conducted through user tests and quantitative evaluations. As a result of learning from this labeled dataset, a network was developed to determine the extrapolation capability of a given video. The proposed classifier achieved much more accurate classification performance than methods that simply use the original video or fixed blur alone by effectively capturing the characteristics of the video through optical flow and adaptive Gaussian blur network. This classifier can be utilized in various fields in conjunction with automatic video extrapolation techniques for immersive viewing experiences.

• Journal of the Korea Computer Graphics Society
• /
• v.30 no.3
• /
• pp.23-41
• /
• 2024

Recently, various virtual avatar music content that showcases singing and dancing have been produced, and as virtual artists gain popularity, offline virtual avatar concerts have also emerged. However, there are few examples of virtual avatar band content where avatars play instruments. In addition, offline virtual avatar concerts using large screens at the front are limited in their ability to utilize the fantastical effects and high degree of freedom unique to virtual reality. In this paper, inspired by these limitations of virtual avatar music content, we introduce the production case of virtual avatar band content and immersive interactive concert of virtual band idol Verse'day. Firstly, we present a case study on creating band performance animations and music videos using motion capture systems and real-time engines. Then, we introduce a production case of an immersive interactive concert using projection mapping technology and a light stick that allows real-time interaction in an offline concert. Finally, based on these production cases, we discussed the future research directions of developing virtual avatar music content creation. We expect that our production cases will inspire the creation of diverse virtual avatar music content and the development of immersive interactive offline virtual avatar concerts in the future.

• Journal of the Korea Computer Graphics Society
• /
• v.30 no.3
• /
• pp.81-86
• /
• 2024

The Image Signal Processor (ISP) converts RAW images captured by the camera sensor into user-preferred sRGB images. While RAW images contain more meaningful information for image processing than sRGB images, RAW images are rarely shared due to their large sizes. Moreover, the actual ISP process of a camera is not disclosed, making it difficult to model the inverse process. Consequently, research on learning the conversion between sRGB and RAW has been conducted. Recently, the ParamISP[1] model, which directly incorporates camera parameters (exposure time, sensitivity, aperture size, and focal length) to mimic the operations of a real camera ISP, has been proposed by advancing the simple network structures. However, existing studies, including ParamISP[1], have limitations in modeling the camera ISP as they do not consider the degradation caused by lens shading, optical aberration, and lens distortion, which limits the restoration performance. This study introduces Positional Encoding to enable the camera ISP neural network to better handle degradations caused by lens. The proposed positional encoding method is suitable for camera ISP neural networks that learn by dividing the image into patches. By reflecting the spatial context of the image, it allows for more precise image restoration compared to existing models.

• Journal of Surface Science and Engineering
• /
• v.57 no.4
• /
• pp.317-324
• /
• 2024

We demonstrate a direct growth of carbon nanotubes (CNTs) on the surface of LiFePO₄ (LFP) powders for use in lithium-ion batteries (LIB). LFP has been widely used as a cathode material due to its low cost and high stability. However, there is a still enough room for development to overcome its low energy density and electrical conductivity. In this study, we fabricated novel structured composites of LFP and CNTs (LFP-CNTs) and characterized the electrochemical properties of LIB. The composites were prepared by direct growth of CNTs on the surface of LFP using a rotary chemical vapor deposition. The growth temperature and rotation speed of the chamber were optimized at 600 ℃ and 5 rpm, respectively. For the LIB cell fabrication, a half-cell was fabricated using polytetrafluoroethylene (PTFE) and carbon black as binder and conductive additives, respectively. The electrochemical properties of LIBs using commercial carbon-coated LFP (LFP/C), LFP with CNTs grown for 10 (LFP/CNTs-10m) and 30 min(LFP/CNTs-30m) are comparatively investigated. For example, after the formation cycle, we obtained 149.3, 160.1, and 175.0 mAh/g for LFP/C, LFP/CNTs-10m, and LFP/CNTs-30m, respectively. In addition, the improved rate performance and 111.9 mAh/g capacity at 2C rate were achieved from the LFP/CNTs-30m sample compared to the LFP/CNTs-10m and LFP/C samples. We believe that the approach using direct growth of CNTs on LFP particles provides straightforward solution to improve the conductivity in the LFP-based electrode by constructing conduction pathways.

• Journal of Korean Society of Archives and Records Management
• /
• v.24 no.3
• /
• pp.85-104
• /
• 2024

While there have been ongoing discussions and attempts to utilize archives for marketing purposes in various organizations, including businesses, there has been a lack of clarity regarding what should be archived and what should be considered as marketing targets within an organization's history. Consequently, historical marketing has often been past-oriented, with results varying significantly based on the capabilities of those in charge. To introduce and effectively utilize archives in organizational settings, it is crucial to demonstrate that archives can positively impact organizational performance. The Heritage Archives is a utilization plan that offers an approach to digitizing and preserving the valuable heritage and assets of a business, explaining them to various stakeholders through records, serving as a foundation for building trust in the business, and linking them to marketing, branding, and other applications. This study focuses on fundamental concepts for constructing and utilizing heritage archives by defining and interpreting key concepts such as the affordance of records, organizational heritage, and heritage assets. To this end, the study incorporates Geoffrey Yeo's affordance and John M.T. Balmer's concept of heritage. In addition, it compares definitions of assets in KS Q ISO 55000:2021, KS X ISO 15489-1:2016, and KS X ISO 30300:2020. Through the study's findings, insights can be obtained for organizations seeking to implement heritage archives and leverage them for marketing, branding, and related purposes.

• Journal of Internet of Things and Convergence
• /
• v.10 no.3
• /
• pp.7-12
• /
• 2024

The development of artificial intelligence technology changes the social structure and educational environment, and the importance of artificial intelligence capabilities continues to increase. This study was conducted with the purpose of developing a checklist of questions to measure AI capabilities of elementary school students. To achieve the purpose of the study, a Delphi survey was used to analyze literature and develop questions. For literature analysis, two domestic studies, five international studies, and the Ministry of Education's curriculum report were collected through a search. The collected data was analyzed to construct core competency measurement elements. The core competency measurement elements consisted of understanding artificial intelligence (6 elements), artificial intelligence thinking (4 elements), artificial intelligence ethics (4 elements), and artificial intelligence social-emotion (3 elements). Considering the knowledge, skills, and attitudes of the constructed measurement elements, 19 questions were developed. The developed questions were verified through the first Delphi survey, and 7 questions were revised according to the revision opinions. The validity of 19 questions was verified through the second Delphi survey. The checklist items developed in this study are measured by teacher evaluation based on performance and behavioral observations rather than a self-report questionnaire. This has the implication that the measurement results of competency are raised to a reliable level.