• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.8
• /
• pp.3942-3961
• /
• 2019

Recently, the models of deep super-resolution networks can successfully learn the non-linear mapping from the low-resolution inputs to high-resolution outputs. However, for large scaling factors, this approach has difficulties in learning the relation of low-resolution to high-resolution images, which lead to the poor restoration. In this paper, we propose Stage Generative Adversarial Networks (Stage-GAN) with semantic maps for image super-resolution (SR) in large scaling factors. We decompose the task of image super-resolution into a novel semantic map based reconstruction and refinement process. In the initial stage, the semantic maps based on the given low-resolution images can be generated by Stage-0 GAN. In the next stage, the generated semantic maps from Stage-0 and corresponding low-resolution images can be used to yield high-resolution images by Stage-1 GAN. In order to remove the reconstruction artifacts and blurs for high-resolution images, Stage-2 GAN based post-processing module is proposed in the last stage, which can reconstruct high-resolution images with photo-realistic details. Extensive experiments and comparisons with other SR methods demonstrate that our proposed method can restore photo-realistic images with visual improvements. For scale factor ${\times}8$, our method performs favorably against other methods in terms of gradients similarity.

• 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.

• Proceedings of the KIEE Conference
• /
• 2003.11b
• /
• pp.295-298
• /
• 2003

In this paper, a novel maximum a posterion (MAP) estimation for the channel decoding of H.264 codes in the presence of transmission error is presented. Arithmetic codes with a forbidden symbol and trellis search techniques are employed in order to estimate the best transmitted. And, there has been growing interest of communication, the research about transmission of exact data is increasing. Unlike the case of voice transmission, noise has a fatal effect on the image transmission. The reason is that video coding standards have used the variable length coding. So, only one bit error affects the all video data compressed before resynchronization. For reasons of that, channel needs the channel codec, which is robust to channel error. But, usual channel decoder corrects the error only by channel error probability. So, designing source codec and channel codec, Instead of separating them, it is tried to combine them jointly. And many researches used the information of source redundancy In received data. But, these methods do not match to the video coding standards, because video ceding standards use not only one symbol but also many symbols in same data sequence. In this thesis, We try to design combined source-channel codec that is compatible with video coding standards. This MAP decoder is proposed by adding semantic structure and semantic constraint of video coding standards to the method using redundancy of the MAP decoders proposed previously. Then, We get the better performance than usual channel coder's.

• Journal of Korean Society of Archives and Records Management
• /
• v.16 no.1
• /
• pp.175-193
• /
• 2016

A topic map was designed as an efficient information retrieval method that is optimized for classification, organization, and navigation through the use of a semantic link network above information resources. With this, this study aims to design a topic map-based university archives retrieval system to provide the relevant information retrieval. For this study, electronic records that relate to the academic administration within two years of D university were collected, and topic map editing was carried out with Ontopia Omnigator. Topics were classified according to their functional analysis of academic administration. In the end, the number of topics was finalized as 626, with 6 types in general: academic work, staff, college register, student, university, etc. Association was separated into six types as well, which were formed with consideration to the relationships among topics. In addition, there are seven occurrence types: register class, register number, register date, receiver, title, creator, and identifier. It is expected that the associative nature of the designed topic map-based retrieval system in this study will make navigation of large records easy and allow incidental discovery of knowledge.

• Journal of Korea Multimedia Society
• /
• v.9 no.12
• /
• pp.1681-1688
• /
• 2006

GIS systems have gradually been utilized in life information as well as special businesses such as traffic, sight-seeing, tracking, and disaster services. Most GIS services focus on showing stored information on maps, not providing a service to register and modify their preferred information. In this paper, we present a new method which reduces DXF map data into Simple Geographic Information File format using format conversion algorithms. We also present the prototype implementation of a GIS search system based on ontologies to support associated information. Our contribution is to propose a new digital map format to provide a fast map loading service and individual customized information on the map service.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.07a
• /
• pp.67-70
• /
• 2022

NLP 분야에서 적극 활용되기 시작한 Transformer 모델을 Vision 분야에서 적용하기 시작하면서 object detection과 segmentation 등 각종 분야에서 기존 CNN 기반 모델의 정체된 성능을 극복하며 향상되고 있다. 또한, label 데이터 없이 이미지들로만 자기지도학습을 한 ViT(Vision Transformer) 모델을 통해 이미지에 포함된 여러 중요한 객체의 영역을 검출하는 saliency map을 추출할 수 있게 되었으며, 이로 인해 ViT의 자기지도학습을 통한 object detection과 semantic segmentation 연구가 활발히 진행되고 있다. 본 논문에서는 ViT 모델 뒤에 classifier를 붙인 모델에 일반 학습한 모델과 자기지도학습의 pretrained weight을 사용해서 전이학습한 모델의 시각화를 통해 각 saliency map들을 비교 분석하였다. 이를 통해, 클래스 분류 학습 기반 전이학습이 transformer의 saliency map에 미치는 영향을 확인할 수 있었다.

• The Journal of Information Systems
• /
• v.30 no.2
• /
• pp.189-214
• /
• 2021

Purpose This paper aims to effectively utilize user-generated content (UGC) and analyze the market structure of a relatively new market which lacks rich user review information. Specifically, we propose a domain-specific text mining tool for the domestic craft beer market and visualize the market structure by incorporating how individual beer products are positioned in the perceptual map of consumers. Design/methodology/approach We collect user review information from Naver blogs, and extract words that describe beers. We identify semantic relationships between beer products through text mining, and then depending on these semantic relationships, construct a graph representing the market structure of the domestic craft beer market based on the consumer's perceptual map. Findings First, beer products produced in the same brewery are perceived as very similar to consumers. Second, only two products, 'Heukdang Milky Stout' and 'Gompyo', was noticeably distinguishable from other products. Third, even though 'Gyeongbokgung' is from a different brewery, it is located very close to the products of 'Jeju Beer' brewery such as 'Jeju Baeknokdam Ale' and 'Seongsan Ilchulbong Ale', which suggests the influence of 'landmark series.' We successfully show that our methodology effectively describes the market structure of the craft beer market.

• The KIPS Transactions:PartB
• /
• v.11B no.7 s.96
• /
• pp.793-802
• /
• 2004

Users in 3D virtual environment get limited information which contains mostly images. It is the main reason for users getting lost during their Navigation. Various studies of Navigation-Aid have been done in order to solve this problem. In this study, we applied Topic Maps, which is one of semantic Web techniques, to the navigation in a 3D virtual environment. Topic Maps construct semantic linking maps through defining the relations between topics. Experiments in which Topic Map based Navigation-Aid was applied have shown that the Navigation-Aid was effective when the subjects find a detailed target rather than a highly represented one. Also, offering information around the target helped the users to find the target when they navigated without having specific targets.

• English Language & Literature Teaching
• /
• v.7 no.1
• /
• pp.137-160
• /
• 2001

This paper is to suggest the way to improve EFL reading skill through semantic mapping by leveled group activities. Semantic mapping is a categorical structuring of information in graphic forms or diagrams. It can be used to activate and organize background knowledge on topics in classrooms. For small group activities, the class is divided into higher leveled groups and lower leveled groups of four members based on their grades. The teaching process has three stages: Pre-reading, while-reading, and post-reading. In the pre-reading stage, students discuss what they know about the topic. They map ideas with a brainstorming technique. In the while-reading stage, they read the text about the topic. While they are reading, they could ask some questions they might have and discuss the information in the text and categorize them with semantic mapping. In the post-reading stage, they discuss what they thought of the topic and add some information about the topic with semantic mapping. For the subjects of this study, third grade, middle school students were selected: 41 students for the experimental group and 35 students for the control group. The experimental period covered almost one semester from March to August, 2000. The results were as follows: 1) The students in the experimental group had higher scores in reading comprehension than those in the control group when semantic mapping was used; 2) The use of semantic mapping in reading comprehension was found to be much more effective in the higher leveled group than in the lower leveled group; 3) The results of questionnaires showed that many students became more interested and motivated in English, and semantic mapping helped them to participate positively in reading the English text. Thus, using semantic mapping by leveled group activities can be an effective alternative to traditional teaching methods for teachers who desire to improve reading skill in middle school students' English classes.

• Journal of KIISE
• /
• v.42 no.9
• /
• pp.1109-1116
• /
• 2015

In this paper, we propose a query processing approach that uses the Spark functional programming and distributed memory system to solve the computational overhead of SPARQL. In the semantic web, RDF ontology data is produced at large scale, and the main challenge for the semantic web is to query and manipulate such a large ontology with a high throughput. The most existing studies on SPARQL have focused on deploying the Hadoop MapReduce framework, and although approaches based on Hadoop MapReduce have shown promising results, they achieve a low level of throughput due to the underlying distributed file processes. Therefore, in order to speed up the query processes, we suggest query- processing methods that are based on memory caching in distributed memory system. Our approach is also integrated with a clause unification method for propagating between the clauses that exploits Spark join, map and filter methods along with caching. In our experiments, we have achieved a high level of performance relative to other approaches. In particular, our performance was nearly similar to that of Sempala, which has been considered to be the fastest query processing system.