• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.140-147
• /
• 2024

In the Republic of Korea, people with hearing impairments are the second-largest demographic within the registered disability community, following those with physical disabilities. Despite this demographic significance, research on sign language translation technology is limited due to several reasons including the limited market size and the lack of adequately annotated datasets. Despite the difficulties, a few researchers continue to improve the performacne of sign language translation technologies by employing the recent advance of deep learning, for example, the transformer architecture, as the transformer-based models have demonstrated noteworthy performance in tasks such as action recognition and video classification. This study focuses on enhancing the recognition performance of sign language translation by combining transformers with 3D-CNN. Through experimental evaluations using the PHOENIX-Wether-2014T dataset [1], we show that the proposed model exhibits comparable performance to existing models in terms of Floating Point Operations Per Second (FLOPs).

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.8
• /
• pp.85-92
• /
• 2024

In this paper presents a comparative analysis of the performance between the latest object detection model, YOLOv10, and its previous versions. YOLOv10 introduces NMS-Free training, an enhanced model architecture, and an efficiency-centric design, resulting in outstanding performance. Experimental results using the COCO dataset demonstrate that YOLOv10-N maintains high accuracy of 39.5% and low latency of 1.84ms, despite having only 2.3M parameters and 6.7G floating-point operations (FLOPs). The key performance metrics used include the number of model parameters, FLOPs, average precision (AP), and latency. The analysis confirms the effectiveness of YOLOv10 as a real-time object detection model across various applications. Future research directions include testing on diverse datasets, further model optimization, and expanding application scenarios. These efforts aim to further enhance YOLOv10's versatility and efficiency.

• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.85-109
• /
• 2018

Recommender system recommends the items expected to be purchased by a customer in the future according to his or her previous purchase behaviors. It has been served as a tool for realizing one-to-one personalization for an e-commerce service company. Traditional recommender systems, especially the recommender systems based on collaborative filtering (CF), which is the most popular recommendation algorithm in both academy and industry, are designed to generate the items list for recommendation by using 'overall rating' - a single criterion. However, it has critical limitations in understanding the customers' preferences in detail. Recently, to mitigate these limitations, some leading e-commerce companies have begun to get feedback from their customers in a form of 'multicritera ratings'. Multicriteria ratings enable the companies to understand their customers' preferences from the multidimensional viewpoints. Moreover, it is easy to handle and analyze the multidimensional ratings because they are quantitative. But, the recommendation using multicritera ratings also has limitation that it may omit detail information on a user's preference because it only considers three-to-five predetermined criteria in most cases. Under this background, this study proposes a novel hybrid recommendation system, which selectively uses the results from 'traditional CF' and 'CF using multicriteria ratings'. Our proposed system is based on the premise that some people have holistic preference scheme, whereas others have composite preference scheme. Thus, our system is designed to use traditional CF using overall rating for the users with holistic preference, and to use CF using multicriteria ratings for the users with composite preference. To validate the usefulness of the proposed system, we applied it to a real-world dataset regarding the recommendation for POI (point-of-interests). Providing personalized POI recommendation is getting more attentions as the popularity of the location-based services such as Yelp and Foursquare increases. The dataset was collected from university students via a Web-based online survey system. Using the survey system, we collected the overall ratings as well as the ratings for each criterion for 48 POIs that are located near K university in Seoul, South Korea. The criteria include 'food or taste', 'price' and 'service or mood'. As a result, we obtain 2,878 valid ratings from 112 users. Among 48 items, 38 items (80%) are used as training dataset, and the remaining 10 items (20%) are used as validation dataset. To examine the effectiveness of the proposed system (i.e. hybrid selective model), we compared its performance to the performances of two comparison models - the traditional CF and the CF with multicriteria ratings. The performances of recommender systems were evaluated by using two metrics - average MAE(mean absolute error) and precision-in-top-N. Precision-in-top-N represents the percentage of truly high overall ratings among those that the model predicted would be the N most relevant items for each user. The experimental system was developed using Microsoft Visual Basic for Applications (VBA). The experimental results showed that our proposed system (avg. MAE = 0.584) outperformed traditional CF (avg. MAE = 0.591) as well as multicriteria CF (avg. AVE = 0.608). We also found that multicriteria CF showed worse performance compared to traditional CF in our data set, which is contradictory to the results in the most previous studies. This result supports the premise of our study that people have two different types of preference schemes - holistic and composite. Besides MAE, the proposed system outperformed all the comparison models in precision-in-top-3, precision-in-top-5, and precision-in-top-7. The results from the paired samples t-test presented that our proposed system outperformed traditional CF with 10% statistical significance level, and multicriteria CF with 1% statistical significance level from the perspective of average MAE. The proposed system sheds light on how to understand and utilize user's preference schemes in recommender systems domain.

• Korean Journal of Ecology and Environment
• /
• v.43 no.3
• /
• pp.428-436
• /
• 2010

This study was to evaluate water quality characteristics and ecological health using a mulimetric fish model in Taehwa-River watershed during May~September 2009. The ecological health assessments were based on the Index of Biological Integrity (IBI) using fish community and the multimetric model of Qualitative Habitat Evaluation Index (QHEI). For the study, the models of IBI and QHEI were modified as 8 and 11 metric attributes, respectively. We also analyzed spatial patterns of chemical water quality over the period of 2000~2009, using the water chemistry dataset, obtained from the Ministry of Environment, Korea. Values of BOD and COD averaged $1.7\;mg\;L^{-1}$ (scope: $0.1{\sim}31.8\;mg\;L^{-1}$) and $3.6\;mg\;L^{-1}$ (scope: $0.4{\sim}33\;mg\;L^{-1}$), respectively during the study. Total nitrogen (TN) and total phosphorus (TP) averaged $2.8\;mg\;L^{-1}$ and $96.8\;{\mu}g\;L^{-1}$, respectively, indicating an eutrophic-hypertrophic state. Also, TN and TP showed longitudinal increases toward the downriver reach. In the watershed, QHEI values varied from 67.5 (fair condition) to 164.5 (good condition) by the criteria of US EPA (1993). There was a abruptly decreasing tendency from T9 site in the QHEI values. According to 1st and 2nd surveys of Taewha River, multimetric model values of IBI was averaged 26.1 (n=14) with "good" condition (B) and the spatial variation was evident. Our results suggest that the mainstream sites was getting worse health condition along the river gradient due to inputs of the point and non-point sources from the urban (Ulsan city). Overall, dataset of IBI, QHEI, and water chemistry indicated that the ecological river health showed a downriver decline and the pattern was closely associated with habitat degradations and chemical pollutions as the waters pass through the urban region.

• Korean Journal of Construction Engineering and Management
• /
• v.17 no.3
• /
• pp.32-42
• /
• 2016

In Korea, the absence of BIM use in existing civil structures and buildings is driving a demand for as-built BIM. As-built BIMs are often created using laser scanners that provide dense 3D point cloud data. Conventional static laser scanning approaches often suffer from limitations in their operability due to the difficulties in moving the equipment, the selection of scanning location, and the requirement of placing targets or extracting tie points for registration of each scanned point cloud. This paper aims at reducing the manual effort using a kinematic 3D laser scanning system based on graph-based simultaneous localization and mapping (SLAM) for continuous indoor mapping. The robotic platform carries three 2D laser scanners: the front scanner is mounted horizontally to compute the robot's trajectory and to build the SLAM graph; the other two scanners are mounted vertically to scan the profiles of surrounding environments. To reduce the accumulated error in the trajectory of the platform through loop closures, the graph-based SLAM system incorporates AdaBoost loop closure approach, which is particularly suitable for the developed multi-scanner system providing more features than the single-scanner system for training. We implemented the proposed method and evaluated it in two indoor test sites. Our experimental results show that the false positive rate was reduced by 13.6% and 7.9% for the two dataset. Finally, the 2D and 3D mapping results of the two test sites confirmed the effectiveness of the proposed graph-based SLAM.

• Korean Journal of Ecology and Environment
• /
• v.40 no.2
• /
• pp.223-233
• /
• 2007

For this study, we analyzed spatial and temporal patterns of trophic state and water quality over the period of $2002{\sim}2005$, using the water chemistry dataset obtained from the Korea Rural community & Agriculture corporation. Most reservoirs, based on TN, showed eutrophic (about 88% of the total). About 20% of agricultural reservoirs, based on TP, showed eutrophic after the criteria of OECD (1982), while 71% and 3% were Hesotrophic and oligotrophic, respectively. Seasonal variations were evident due to the intense monsoon rain during July${\sim}$August; conductivity, COD, SS, nutrients, and chlorophyll-${\alpha}$ (CBL) increased in the postmonsoon compared to the premonsoon. TP values had positive functional relations with conductivity, COD, and CHL values. COD and SS peaked during the intense monsoon. Mean values of TP and CHL values were two times greater in the intense monsoon than the weak monsoon. The increased TP was probably due to inorganic suspended solids from point and non-point sources during the monsoon. Ratios of TN : TP had strong in- verse relations ($R^2$=0.843, p<0.001, n=34) with TP, but not with TN (p>0.05, n=34). Log10-transformed CHL increased with TP in most P-limited reservoirs $(Log_{10}TP=0.5{\times}Log_{10}CHL+0.086)$. Similarity analysis, based TN, TP, and CHL showed that three groups were separated at 90% similarity level; One group was reservoirs with high salinity nearby the seawater, and the other two groups were reservoirs with a low salinity of the inland, and intermediate salinity, respectively.

• Korean Journal of Remote Sensing
• /
• v.29 no.1
• /
• pp.21-34
• /
• 2013

Permanent Scatterer InSAR (PSInSAR) technique extracts permanent scatterers exhibiting high phase stability over the entire observation period and calculates precise time-series deformation at Permanent Scatterer (PS) points by using single master interferograms. This technique is not a good method to apply on nature environment such as forest area where permanent scatterers cannot be identified. Another muti-temporal Interferometric Synthetic Aperture Radar (InSAR), Small BAseline Subset (SBAS) technique using multi master interferograms with short baselines, can be effective to detect deformation in forest area. However, because of the error induced from phase unwrapping, the technique sometimes fails to estimate correct deformation from a stack of interferograms. To overcome those problems, we introduced new multi-temporal InSAR technique, called Temporarily Coherence Point InSAR (TCPInSAR), in this paper. This technique utilizes multi master interferograms with short baseline and without phase unwrapping. To compare with traditional multi-temporal InSAR techniques, we retrieved spatially changing deformation because PSs have been found enough in forest area with TCPInSAR technique and time-series deformation without phase unwrapping error. For this study, we acquired ERS-1 and ERS-2 SAR dataset on Augustine volcano, Alaska and detected deformation in study area for the period 1992-2005 with SBAS and TCPInSAR techniques.

• Korean Journal of Remote Sensing
• /
• v.30 no.5
• /
• pp.571-586
• /
• 2014

Moderate Resolution Imaging Spectroradiometer (MODIS) provides air temperature (Tair) and dew point temperature (Tdew) profiles at a spatial resolution of 5 km. New Collection 6 (C006) MODIS07_L2 atmospheric profile product has been produced since 2012. The Collection 6 algorithm has several modifications from the previous Collection 5 (C005) algorithm. This study evaluated reliabilities of two alternative datasets of surface-level Tair and Tdew derived from C005 and C006 Aqua MODIS07_L2 (MYD07_L2) products using ground measured temperatures from 77 National Weather Stations (NWS). Saturated and actual vapor pressures were calculated using MYD07_L2 Tair and Tdew. The C006 Tair showed lower mean error (ME, -0.76 K) and root mean square error (RMSE, 3.34 K) than the C005 Tair (ME = -1.89 K, RMSE = 4.06 K). In contrasts, ME and RMSE of C006 Tdew were higher than those (ME = -0.39 K, RMSE = 5.65 K) of C005 product. Application of ambient lapse rate for Tair showed appreciable improvements of estimation accuracy for both of C005 and C006, though this modification slightly increased errors in C006 Tdew. The C006 products provided better estimation of vapor pressure datasets than the C005-derived vapor pressure. Our results indicate that, except for Tdew, C006 MYD07_L2 product showed better reliability for the region of South Korea than the C005 products.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.8
• /
• pp.23-30
• /
• 2021

In this study, we present a comparative analysis of major autoencoder(AE)-based anomaly detection methods for quality determination in the manufacturing process and a new anomaly discrimination criterion. Due to the characteristics of manufacturing site, anomalous instances are few and their types greatly vary. These properties degrade the performance of an AI-based anomaly detection model using the dataset for both normal and anomalous cases, and incur a lot of time and costs in obtaining additional data for performance improvement. To solve this problem, the studies on AE-based models such as AE and VAE are underway, which perform anomaly detection using only normal data. In this work, based on Convolutional AE, VAE, and Dilated VAE models, statistics on residual images, MSE, and information entropy were selected as outlier discriminant criteria to compare and analyze the performance of each model. In particular, the range value applied to the Convolutional AE model showed the best performance with AUC PRC 0.9570, F1 Score 0.8812 and AUC ROC 0.9548, accuracy 87.60%. This shows a performance improvement of an accuracy about 20%P(Percentage Point) compared to MSE, which was frequently used as a standard for determining outliers, and confirmed that model performance can be improved according to the criteria for determining outliers.

• Explosives and Blasting
• /
• v.40 no.3
• /
• pp.19-32
• /
• 2022

Hydrogen is a fuel having the highest energy compared with other common fuels. This means hydrogen is a clean energy source for the future. However, using hydrogen as a fuel has implication regarding carrier and storage issues, as hydrogen is highly inflammable and unstable gas susceptible to explosion. Explosions resulting from hydrogen-air mixtures have already been encountered and well documented in research experiments. However, there are still large gaps in this research field as the use of numerical tools and field experiments are required to fully understand the safety measures necessary to prevent hydrogen explosions. The purpose of this present study is to develop and simulate 3D numerical modelling of an existing hydrogen gas station in Jeonju by using handheld LiDAR and Ansys AUTODYN, as well as the processing of point cloud scans and use of cloud dataset to develop FEM 3D meshed model for the numerical simulation to predict peak-over pressures. The results show that the Lidar scanning technique combined with the ANSYS AUTODYN can help to determine the safety distance and as well as construct, simulate and predict the peak over-pressures for hydrogen refueling station explosions.