• The Korean Journal of Applied Statistics
• /
• v.34 no.2
• /
• pp.191-204
• /
• 2021

Visual Question Answering (VQA) and image captioning are tasks that require understanding of the features of images and linguistic features of text. Therefore, co-attention may be the key to both tasks, which can connect image and text. In this paper, we propose a model to achieve high performance for VQA by image caption generated using a pretrained standard transformer model based on MSCOCO dataset. Captions unrelated to the question can rather interfere with answering, so some captions similar to the question were selected to use based on a similarity to the question. In addition, stopwords in the caption could not affect or interfere with answering, so the experiment was conducted after removing stopwords. Experiments were conducted on VQA-v2 data to compare the proposed model with the deep modular co-attention network (MCAN) model, which showed good performance by using co-attention between images and text. As a result, the proposed model outperformed the MCAN model.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.3
• /
• pp.529-536
• /
• 2022

In the past two years, Severe Acute Respiratory Syndrome Coronavirus-2(SARS-CoV-2) has been hitting more and more to people. This paper proposes a novel U-Net Convolutional Neural Network to classify and segment COVID-19 lung CT images, which contains Sub Coding Block (SCB), Atrous Spatial Pyramid Pooling(ASPP) and Attention Gate(AG). Three different models such as FCN, U-Net and U-Net-SCB are designed to compare the proposed model and the best optimizer and atrous rate are chosen for the proposed model. The simulation results show that the proposed U-Net-MMFE has the best Dice segmentation coefficient of 94.79% for the COVID-19 CT scan digital image dataset compared with other segmentation models when atrous rate is 12 and the optimizer is Adam.

• The KIPS Transactions:PartB
• /
• v.17B no.1
• /
• pp.47-54
• /
• 2010

Old archived film shows two major defects: line scratch and blobs. In this paper, we present a design and implementation of an automatic video restoration system for line scratches observed in archived film. We use autoregressive (AR) image model because we can make stochastic and specifically autoregressive image generation process with our PAST-PRESENT model and Sampling Pattern. We designed locality maximizing scanning pattern, which can generate nearly stationary time-like series of pixels, which is a strong requirement for a stochastic series to be autoregressive. The sampled pixel series undergoes filtering and model fitting using Durbin-Levinson algorithm before interpolation process. We designed three-stage film restoration system, which includes (1) film acquisition from VHS tapes, (2) simple line scratch detection and restoration, and (3) manual blob identification and sophisticated inpainting scheme. We implemented film acquisition and simple inpainting scheme on Texas Instruments DSP board TMS320DM642 EVM, and implemented our AR inpainting scheme on PC for sophisticated restoration. We experimented our scheme with two old Korean films: "Viva Freedom" and "Robot Tae-Kwon-V", and the experimental results show that our scheme improves Bertalmio's scheme for subjective quality (MOS), objective quality (PSNR), and especially restoration ratio (RR), which reflects how much similar to the manual inpainting results.

• Journal of KIISE
• /
• v.44 no.1
• /
• pp.84-94
• /
• 2017

Recently, the importance of velocity, one of the characteristics of big data (5V: Volume, Variety, Velocity, Veracity, and Value), has been emphasized in the data processing, which has led to several studies on the real-time stream processing, a technology for quick and accurate processing and analyses of big data. In this paper, we propose a Squall framework using Time-triggered Message-triggered Object (TMO) technology, a model that is widely used for processing real-time big data. Moreover, we provide a description of Squall framework and its operations under a single node. TMO is an object model that supports the non-regular real-time processing method for certain conditions as well as regular periodic processing for certain amount of time. A Squall framework can support the real-time event stream of big data and micro-batch processing with outstanding performances, as compared to Apache storm and Spark Streaming. However, additional development for processing real-time stream under multiple nodes that is common under most frameworks is needed. In conclusion, the advantages of a TMO model can overcome the drawbacks of Apache storm or Spark Streaming in the processing of real-time big data. The TMO model has potential as a useful model in real-time big data processing.

• Korean Journal of Construction Engineering and Management
• /
• v.16 no.3
• /
• pp.91-100
• /
• 2015

An increase in energy such as natural gas, coal, oil, has occurred to a large amounts of environment impact emissions, it is necessary to reduce in the construction industry for the energy consumption. To encourage remodeling project in developed countries of the majority, on the basis of this, remodeling project in the construction industry has grown to a large amount. Results of analysis of the research related to the advanced remodeling, analysis of the economic validity in accordance with the production and process and building elapsed years of selection alternative of remodeling there has been a problem that has not been properly reflected. In this study, a decision support model that can simultaneously choose the most cost-effective and energy-efficiency alternative. Developed process model, generates a "Remodeling Solution" that combines the renewable energy equipment and envelope system, energy performance evaluation of the application of international standards(ISO-13790, DIN V 18599), perform the economic evaluation through LCCA(Life Cycle Cost Analysis) technique, circulated evaluation and configured to output the optimal Remodeling Solution. The results of applying the model developed in the case, it was confirmed that it is possible to select a choice of cost-effective energy-saving alternative. Then, developed model through this study, it is expected to be able to help highly effective remodeling alternative to selecting by decision-makers.

• The Korean Journal of Pesticide Science
• /
• v.11 no.2
• /
• pp.72-81
• /
• 2007

Quantitative structure-activity relationships (QSARs) on the pre-emergency herbicidal activity and reactivity of a series of new O,O-dialkyl-1-phenoxyacetoxy-1-methylphosphonates (S) analogues against seed of cucumber (Cucumus Sativa) were discussed quantitatively using 2D-QSAR and HQSAR methods. The statistical values of HQSAR model were better than that of 2D-QSAR model. From the frontier molecular orbital (FMO) interaction between substrate molecule (S) and $BH^+$ ion (I) in PDH enzyme, the electrophilic reaction was superior in reactivity. From the effect of substituents, $R_2$-groups in substrate molecule (S) contributed to electrophilic reaction with carbonyl oxygen atom while X, Y-groups contributed to nucleophilic reaction with carbonyl carbon atom. And the influence of X,Y-groups was more effective than that of $R_2$-groups. As a results of 2D-QSAR model (I & II) and atomic contribution maps with HQSAR model, the more length of X, Y-groups is longer, the more herbicidal activity tends to increased. And also, the optimal ${\epsilon}LUMO$ energy, $({\epsilon}LUMO)_{opt.}$=-0.479 (e.v.) of substrate molecule is important factor in determining the herbicidal activity. It is predicted that the herbicidal activity proceeds through a nucleophilic reaction. From the analytical results of 2D-QSAR and HQSAR model, it is suggested that the structural distinctions and descriptors that contribute to herbicidal activities will be able to applied new herbicide design.

• Korean Journal of Remote Sensing
• /
• v.40 no.1
• /
• pp.71-80
• /
• 2024

In this study, a proposed ensemble learning technique aims to enhance the semantic segmentation performance of images captured by Unmanned Aerial Vehicles (UAVs). With the increasing use of UAVs in fields such as urban planning, there has been active development of techniques utilizing deep learning segmentation methods for land cover segmentation. The study suggests a method that utilizes prominent segmentation models, namely U-Net, DeepLabV3, and Fully Convolutional Network (FCN), to improve segmentation prediction performance. The proposed approach integrates training loss, validation accuracy, and class score of the three segmentation models to enhance overall prediction performance. The method was applied and evaluated on a land cover segmentation problem involving seven classes: buildings,roads, parking lots, fields, trees, empty spaces, and areas with unspecified labels, using images captured by UAVs. The performance of the ensemble model was evaluated by mean Intersection over Union (mIoU), and the results of comparing the proposed ensemble model with the three existing segmentation methods showed that mIoU performance was improved. Consequently, the study confirms that the proposed technique can enhance the performance of semantic segmentation models.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.6
• /
• pp.392-399
• /
• 2023

This paper proposes a deep learning model to determine the region and depth of cabbage cores for robotic automation of the cabbage core removal process during the kimchi manufacturing process. In addition, rather than predicting the depth of the measured cabbage, a model was presented that simultaneously detects and classifies the area by converting it into a discrete class. For deep learning model learning and verification, RGB images of the harvested cabbage 522 were obtained. The core region and depth labeling and data augmentation techniques from the acquired images was processed. MAP, IoU, acuity, sensitivity, specificity, and F1-score were selected to evaluate the performance of the proposed YOLO-v4 deep learning model-based cabbage core area detection and classification model. As a result, the mAP and IoU values were 0.97 and 0.91, respectively, and the acuity and F1-score values were 96.2% and 95.5% for depth classification, respectively. Through the results of this study, it was confirmed that the depth information of cabbage can be classified, and that it can be used in the development of a robot-automation system for the cabbage core removal process in the future.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.1
• /
• pp.317-326
• /
• 2024

Biometric recognition is a technology that determines whether a person is identified by extracting information on a person's biometric and behavioral characteristics with a specific device. Cyber threats such as forgery, duplication, and hacking of biometric characteristics are increasing in the field of biometrics. In response, the security system is strengthened and complex, and it is becoming difficult for individuals to use. To this end, multiple biometric models are being studied. Existing studies have suggested feature fusion methods, but comparisons between feature fusion methods are insufficient. Therefore, in this paper, we compared and evaluated the fusion method of multiple biometric models using fingerprint, face, and iris images. VGG-16, ResNet-50, EfficientNet-B1, EfficientNet-B4, EfficientNet-B7, and Inception-v3 were used for feature extraction, and the fusion methods of 'Sensor-Level', 'Feature-Level', 'Score-Level', and 'Rank-Level' were compared and evaluated for feature fusion. As a result of the comparative evaluation, the EfficientNet-B7 model showed 98.51% accuracy and high stability in the 'Feature-Level' fusion method. However, because the EfficietnNet-B7 model is large in size, model lightweight studies are needed for biocharacteristic fusion.

• Korean Journal of Fisheries and Aquatic Sciences
• /
• v.30 no.5
• /
• pp.782-793
• /
• 1997

To study the physiological adaptation (shift-up) of phytoplankton under the simulated upwelling conditions, nitrate uptake capacity of Dunaliella tertiolecta batch culture was measured in the laboratory using the stable isotope $^{15}N-KNO_3$. Contrary to the expected, there was no significant relationship between the maximum $V_{NO3}$ (nitrogen specific nitrate uptake rate) and the initial nitrate concentration. However, there was a strong relationship between the maximum $\rho_{NO3}$ (nitrate transport rate) and the initial nitrate concentration of $<25\;{\mu}M$, which was also influenced by the physiological status of the culture. The increase in $V_{NO3}$ was mainly due to the increase in PON (particulate organic nitrogen) concentration and partly due to the increase in $V_{NO3}$. When the phytoplankton population was severely shifted-down, the physiological adaptation of nitrate uptake was significantly inhibited at high initial nitrate concentrations. The timing of the maximum $V_{NO3}$ or $\rho_{NO3}$ was related to the initial nitrate concentration. At higher initial nitrate concentrations, maxima in $V_{NO3}$ and $\rho_{NO3}$ occurred 1 or 2 days later than at lower nitrate concentrations. This relationship was the opposite to the prediction from the shift-up model of Zimmerman et al. (1987), The shift-up process is apparently controlled by an internal time sequence and the initial nitrate concentration, but the magnitude of $V_{NO3}$ was affected little by changes in nitrate concentration.