• ETRI Journal
• /
• v.45 no.5
• /
• pp.822-835
• /
• 2023

Recent semantic segmentation frameworks usually combine low-level and high-level context information to achieve improved performance. In addition, postlevel context information is also considered. In this study, we present a Context ReFinement Network (CRFNet) and its training method to improve the semantic predictions of segmentation models of the encoder-decoder structure. Our study is based on postprocessing, which directly considers the relationship between spatially neighboring pixels of a label map, such as Markov and conditional random fields. CRFNet comprises two modules: a refiner and a combiner that, respectively, refine the context information from the output features of the conventional semantic segmentation network model and combine the refined features with the intermediate features from the decoding process of the segmentation model to produce the final output. To train CRFNet to refine the semantic predictions more accurately, we proposed a sequential training scheme. Using various backbone networks (ENet, ERFNet, and HyperSeg), we extensively evaluated our model on three large-scale, real-world datasets to demonstrate the effectiveness of our approach.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.6
• /
• pp.319-325
• /
• 2023

In this study, we proposed a method for hyperparameter optimization in the building and training of a deep learning model designed to process point cloud data collected by a millimeter-wave radar system. The primary aim of this study is to facilitate the deployment of a baseline model in resource-constrained IoT devices. We evaluated a RadHAR baseline deep learning model trained on a public dataset composed of point clouds representing five distinct human activities. Additionally, we introduced a coarse-to-fine hyperparameter optimization procedure, showing substantial potential to enhance model efficiency without compromising predictive performance. Experimental results show the feasibility of significantly reducing model size without adversely impacting performance. Specifically, the optimized model demonstrated a 3.3% improvement in classification accuracy despite a 16.8% reduction in number of parameters compared th the baseline model. In conclusion, this research offers valuable insights for the development of deep learning models for resource-constrained IoT devices, underscoring the potential of hyperparameter optimization and model size reduction strategies. This work contributes to enhancing the practicality and usability of deep learning models in real-world environments, where high levels of accuracy and efficiency in data processing and classification tasks are required.

• Microbiology and Biotechnology Letters
• /
• v.45 no.3
• /
• pp.236-242
• /
• 2017

The conversion of marine biomass to renewable energy has been considered an alternative to fossil fuels. Butanol, in particular, can be used directly as a fuel. In this experiment, the brown alga Undaria pinnatifida was selected as a biomass for biobutanol production. Hyper thermal (HT) acid hydrolysis was used as an acid hydrolysis method to produce monosaccharides. The optimal pretreatment conditions for U. pinnatifida were determined as slurry with 10% (w/v) U. pinnatifida content and 270 mM $H_2SO_4$, and heating at $160^{\circ}C$ for 7.5 min. Enzymatic saccharification was carried out with Celluclast 1.5 L, Viscozyme L, and Ultraflo Max. The optimal saccharification condition was 12 U/ml Viscozyme L. Fermentations were carried out for the production of acetone, butanol, and ethanol by Clostridium acetobutylicum KCTC 1724, Clostridium beijerinckii KCTC 1785, and Clostridium tyrobutyricum KCTC 5387. The fermentations were carried out using a pH-control. The optimal ABE fermentation condition determined using C. acetobutylicum KCTC 1724 adapted to 160 g/l mannitol. An ABE concentration of 9.05 g/l (0.99 g/l acetone, 5.62 g/l butanol, 2.44 g/l ethanol) was obtained by the consumption of 24.14 g/l monosaccharide with $Y_{ABE}$ of 0.37 in pH 5.0.

• The Journal of Korean Physical Therapy
• /
• v.15 no.4
• /
• pp.112-128
• /
• 2003

Since areas of pain and dysfunction of musculoskeletal typically suffered by the patients with back disorders spread all over the body, WBIP(GCM Program) for the primary treatment and management is required. The purpose of this study is to analyze if WBIP(GCM Program) based on the hyper/hypomobility pattern of Four Body Types can identify the effective treatment of back disorders and the effect on the postural balanced restoration of the spine and extremities. Non-specific back disorder is still a major reason for sick leave. And moreover, its been reported that there was often recurrence to the patients whose symptom had been diminished. As a WBIP(GCM Program) based on kinematic chain patterns of Four Body Types, this study gave a new information on the effective diagnosis, treatment and management of non-specific back disorders. 337 patients above the twenty-five years old with the non-specific back disorders at the hospital and oriental medical clinics at Kyungnam and Busan areas in South Korea from August 24th, 2000 to Feb 23rd, 2001 have randomly been assigned to four experimental groups such as Whole Body Intervention Program Group, Physical Therapy Group like modality treatments, Acupuncture-Treatment Group, and Placebo Control Group. According to intervention program applied to the each four group for three times per week(twelve times per 4weeks), as the time-series methods, we compared and evaluated the body status of the pretest with that of post treatment completion of four week, three month, and six month, respectively. As the analytical method of measurement, our researchers used the Moire Interferometry Unit and Postural Kit that could measure the postural balance of spine and extremities. The collection of data was performed in the designated hospital and oriental medical clinics. For the analysis of the data, the SPSS 10.0 package program was used. X2-test has been taken in order to compare and analyze characteristics and GPES of the patients in four experimental groups. Repeated Measure ANOVA and Tukey post hoc test has been adopted in order to compare the effects of the balanced restoration of the spine and extremities among four Groups categorized for this study. Statistical significance was accepted at the 0.05 level of confidence The effect of the balanced restoration on the spine and extremities of the patients with non-specific back disorders has been proved in all of the Groups. As for the restoration degree, however, WBIP(GCM Program) Group produced the highest effectiveness in terms of the fact that it had a dense moire in comparison with the other three Groups and that the Moires of both sides had the same level by the time(p<0.01). WBIP(GCM Program) based on four tilting types of scapular and ilium and hyper/hypomobility pattern took a higher effect on the balanced restoration of the spine and extremities through a whole body as well as the treatment of back disorders than the other three Groups which the usual remedy without classification of body type had been applied to.

• Journal of Society of Preventive Korean Medicine
• /
• v.13 no.1
• /
• pp.41-58
• /
• 2009

Objectives: The purpose of this study was to investigate the effects of Hyolbuchukeo-tang extracts on the hyperlipidemia rats induced by high fat diet. Methods and Materials: In vitro; The extracts prepared for Hyolbuchukeo-tang by hot water extraction (HH), fermentation(HF) and UMPM extraction(HU) method. The extracts were examined for levels of polyphenol compounds, antioxidant activities, and inhibitory potencies for HMG-CoA reductase. In vivo; Sprague-Dawley male rats of weighing $150{\pm}5g$ were randomly divided into six groups ; normal control diet(NC), and high fat diet(HC), high fat diet with treated lovastatin of 10mg/kg(PC), high fat diet with Hyolbuchukeo-tang extracts; HH, HF and HU treated extracts of 300mg/kg, respectively. Also, we compared the effects of the extracts of HH, HF and HU on rats fed high fat diet for four weeks. Results: 1. The content of polyphenol compounds and electron donating abilities(EDAs) was the HF higher than HH and HU. The superoxide dismutase(SOD)-like activities were proportionate in consistency and they appeared highly from all extracts. The HMG-CoA reductase inhibition activities was highest activities in the HU. 2. The activities of serum GOT and GPT were significantly lower in the HH and HF groups. The level of serum triglyceride was significantly decreased in the HF group. HH and HU groups were significantly decreased in the atherogenic index(AI). The total cholesterol concentration in liver was significantly decreased in the HF group, and HU showed more significantly decreased in the triglyceride than of the lovastatin. Also, photomicrographs of liver tissue showed higher fat accumulation in the HC group than in the HH, HF and HU groups. Conclusions: These result suggest that the hyper-lipidemia caused by a high fat diet was effectively inhibited the administration of HF and HU.

• Journal of KIISE:Computing Practices and Letters
• /
• v.7 no.2
• /
• pp.165-182
• /
• 2001

In this paper, we describe a design and implementation of the Wireless Markup Language(WML) compiler to deploy wireless Internet services effectively. The WML compiler translates textual WML decks into binary ones in order to reduce the traffic on wireless links that have relatively low bandwidth to wireline links and mitigate the processing overhead of WML decks on, wireless terminals that have relatively low processing power to fixed workstations. In addition, it takes over the overhead of eXtensible Markup Language(XML) well-formedness and validation processes. The WML compiler consists of the lexical analyzer and parser modules. The granunar for the WML parser module is LALR(1) context-free grammar that is designed based on XML 1.0 and WML 1.2 DTD(Document Type Definition) with the consideration of the Wireless Application Protocol Binary XML grammar. The grammar description is converted into a C program to parse that grammar by using parser generator. Even though the tags in WML will be extended or WML DTD will be upgraded, this approach has the advantage of flexibility because the program is generated by modifying just the changed parts. We have verified the functionality of the WML compiler by using a WML decompiler in the public domain and by using the Nokia WAP Toolkit as a WAP client. To measurethe compressibility gain of the WML compiler, we have tested a large number of textual WML decks and obtained a maximum 85 %. As the effect of compression is reduced when the portion of general textual strings increases relative to one of the tags and attributes in a WML deck, an extended encoding method might be needed for specific applications such as compiling of the WML decks to which the Hyper Text Markup Language document is translated dynamically.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.5
• /
• pp.150-164
• /
• 2013

Currently it is being gradually focused on the mobile application's processing performance implemented by Javascript and HTML (Hyper Text Markup Language) due to the dissemination of mobile web application supply based on the WAC (Wholesale Application Community). If the application software has a simple functional processing structure, then the problem is benign, however, the load of a browser is getting heavier as the amount of Javascript processing is being increased. There is a limitation on the processing time and capacity of the Javascript in the ordinary mobile browsers which are on the market now. In order to solve those problems, the Web Worker that is not supported from the existing Javascript technology is now provided by the HTML 5 to implement the multi thread. The Web Worker provides a mechanism that process a part from the single thread through a separate one. However, it can not guarantee the computing ability as a native application on the mobile and is not enough as a solution for improving the fundamental processing speed. The Cyclostorm overcomes the limitation of resources as a mobile client and guarantees the performance as a native application by providing high computing service and ascripting the Javascript process on the mobile to the computer server on the cloud. From the performance evaluation experiment, the Cyclostorm shows a maximally 6 times faster computing speed than in the existing mobile browser's Javascript and 3 to 6 times faster than in Web Worker of the HTML 5. In addition, the usage of memory is measured less than the existing method since the server's memory has been used. In this paper, the Cyclostorm is introduced as one of the mobile cloud computing services to conquer the limitation of the WAC based mobile browsers and to improve the existing web application's performances.

• Journal of Navigation and Port Research
• /
• v.44 no.3
• /
• pp.187-194
• /
• 2020

It is very important to forecast freight volume accurately to establish major port policies and future operation plans. Thus, related studies are being conducted because of this importance. In this paper, stepwise regression analysis and artificial neural network model were analyzed to compare the predictive power of each model on Gwangyang Port, the largest domestic port for coal and iron ore transportation. Data of a total of 121 months J anuary 2009-J anuary 2019 were used. Factors affecting coal and iron ore trade volume were selected and classified into supply-related factors and market/economy-related factors. In the stepwise regression analysis, the tonnage of ships entering the port, coal price, and dollar exchange rate were selected as the final variables in case of the Gwangyang Port coal volume forecasting model. In the iron ore volume forecasting model, the tonnage of ships entering the port and the price of iron ore were selected as the final variables. In the analysis using the artificial neural network model, trial-and-error method that various Hyper-parameters affecting the performance of the model were selected to identify the most optimal model used. The analysis results showed that the artificial neural network model had better predictive performance than the stepwise regression analysis. The model which showed the most excellent performance was the Gwangyang Port Coal Volume Forecasting Artificial Neural Network Model. In comparing forecasted values by various predictive models and actually measured values, the artificial neural network model showed closer values to the actual highest point and the lowest point than the stepwise regression analysis.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.140-142
• /
• 2021

Mobile communication technology after 5th generation requires high speed, hyper-connection, and low latency communication. In order to meet technical requirements for secure hyper-connectivity, low-spec IoT devices that are considered the end of IoT services must also be able to provide the same level of security as high-spec servers. For the purpose of performing these security functions, it is required for cryptographic keys to have the necessary degree of stability in cryptographic algorithms. Cryptographic keys are usually generated from cryptographic random number generators. At this time, good noise sources are needed to generate random numbers, and hardware random number generators such as TRNG are used because it is difficult for the low-spec device environment to obtain sufficient noise sources. In this paper we used the chip which is based on quantum characteristics where the decay of radioactive isotopes is unpredictable, and we presented a variety of methods (TRNG) obtaining an entropy source in the form of binary-bit series. In addition, we conducted the NIST SP 800-90B test for the entropy of output values generated by each TRNG to compare the amount of entropy with each method.

• The Journal of the Korea Contents Association
• /
• v.8 no.10
• /
• pp.54-63
• /
• 2008

With the improvement of computer performance in Mobile Environments, the users demand for the multimedia player is getting higher, which will be the major force behind the wide distribution of the media player that allows several moving pictures to run simultaneously using multiple windows on the screen. Products that adopted the same method used for a single moving picture and yet support the several moving pictures simultaneously emerged; however, they are not immune from problems. In this paper, we propose method to solve the problems of blinking, play prioritization processor and semi-transparent processor that occur when appling several moving pictures to the method used for a single moving picture and suggested the possible solutions. Taking it one step further, it focused on design and implementation of the media player that allows several moving pictures using overlapping and overlay technique by recognizing the biggest problem, which is that the above media play is not supported when playing several moving pictures simultaneously.