• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.853-855
• /
• 2021

본 연구는 회화문화재 속에 표현된 다양한 종류의 객체를 검출할 수 있는 딥러닝 모델생성을 위해 필요한 학습용 이미지 데이터셋 구축방안을 제안한다. 먼저 기존 동양화 기반의 회화문화재 이미지 데이터 및 객체 특징 분석을 진행하였고, 이를 바탕으로 Natural image에 Pose transfer 및 Style transfer를 적용한 새로운 방식의 회화문화재 이미지 데이터 생성 방법을 제안한다. 제안한 프레임워크를 통해 기존 문화재 분야에서 가지고 있던 제한된 데이터 구축문제를 극복하고, 검출모델 생성을 위한 대용량의 학습데이터 구축 가능성을 제시하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.562-565
• /
• 2021

스타일 변환 모델은 이미지 전체나 이미지 내에서 사용자가 지정한 영역을 대상으로 스타일을 변환시킨다. 이런 방식은 이미지 내의 다수의 객체에 대해 스타일 변환을 시행할 때 일일이 영역을 지정해 줘야 한다는 불편함과 결과물의 전체 해상도가 떨어진다는 한계를 가지고 있다. 본 논문에서는 이런 한계들을 극복하기 위해 객체탐지 모델과 스타일변환 모델을 연동한 객체스타일변환모델을 제안하고 모델 간 연동방법에 대해 자세히 서술한다. 객체탐지모델인 Mask R-CNN 을 통해 필요한 객체를 탐지하고 탐지한 객체의 특징맵들을 스타일변환 모델인 zi2zi 의 입력 값으로 전달하여 이미지 내의 필요한 객체들만 스타일변환이 이루어지도록 모델이 동작한다. 이러한 모델은 기존에 있는 두 모델을 재사용함으로써 모델을 처음부터 새로 설계할 필요가 없다는 장점이 있으며, 공개된 다양한 모델들을 서로 융합하여 사용할 수 있는 방법을 제시하는데 도움을 줄 것이다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.329-332
• /
• 2022

본 연구에서는 하나의 2D 이미지를 StyleGAN을 통해 다각도의 이미지를 생성하고, 그것을 다시 Kaolin으로 구현한 역그래픽 렌더러의 입력으로 받아 3D 오브젝트로 변환한다. 또한, 클레이튼 기반의 블록체인을 통해 NFT 기술을 통하여 3D 오브젝트를 NFT로 만들 수 있도록 한다. 최종적으로 2D 이미지를 메타버스에서 활용할 수 있는 3D 패션 아이템으로 변환하고 NFT를 발행하여 거래한다. 본 연구는 개인이 자유롭게 메타버스 콘텐츠를 제공하고 거래하여 메타버스 활성화에 기여할 것으로 기대한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.11a
• /
• pp.1087-1090
• /
• 2008

Web 경계를 허무는 새 시대에 맞추어 라이프스타일은 급속히 변화하고 있다. 가상의 3D 복합 서비스는 현실과 다름없는 또 하나의 삶을 제공함으로써, 가상세계와 현실세계의 경계를 없애고 있고 이전 많은 사람들이 제안하고 예상했던, 현실의 정보가 가상세계에서 또한 다름 없는 데이터로 존재함으로써 가상세계와 현실세계 사이의 벽을 허물고 그 차이를 느끼리 못 하도록 하는 연구를 하고자 하였다. 우선 손쉽게 제시할 수 있는 실험을 통하여 그 가능성을 판단해 보고자 하였으며, 간단한 예로써 현실세계를 대표할 모바일 장비와 가상 현실세계의 대표적인 예인 SecondLife 두 가지 시스템의 접목을 통하여 가상현실 세계와 현실간의 데이터 공유 환경을 만들고 정보를 공유를 통하여, 얻을 수 있는 장점들과 활용방안에 대하여 찾고자 한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.794-797
• /
• 2024

Despite advancements in deep learning, existing semantic segmentation models exhibit suboptimal performance under adverse weather conditions, such as fog or rain, whereas they perform well in clear weather conditions. To address this issue, much of the research has focused on making image or feature-level representations weather-independent. However, disentangling the style and content of images remains a challenge. In this work, we propose a novel fine-tuning method, 'freeze-n-update.' We identify a subset of model parameters that are weather-independent and demonstrate that by freezing these parameters and fine-tuning others, segmentation performance can be significantly improved. Experiments on a test dataset confirm both the effectiveness and practicality of our approach.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.2-5
• /
• 2022

With the advent of deep learning and the rapid development of ICT (Information and Communication Technology), research using artificial intelligence is being actively conducted in various fields of society such as politics, economy, and culture and so on. Deep learning-based artificial intelligence technology is subdivided into various domains such as natural language processing, image processing, speech processing, and recommendation system. In particular, as the industry is advanced, the need for a recommendation system that analyzes market trends and individual characteristics and recommends them to consumers is increasingly required. In line with these technological developments, this paper extracts and classifies attribute information from structured or unstructured text and image big data through deep learning-based technology development of 'language processing intelligence' and 'image processing intelligence', and We propose an artificial intelligence-based 'customized fashion advisor' service integration system that analyzes trends and new materials, discovers 'market-consumer' insights through consumer taste analysis, and can recommend style, virtual fitting, and design support.

• Journal of the Korea Computer Graphics Society
• /
• v.23 no.2
• /
• pp.11-22
• /
• 2017

This study proposes IMToon(IMage-based carToon) which is an image-based cartoon authoring system using an image processing algorithm. The proposed IMToon allows general users to easily and efficiently produce frames comprising cartoons based on image. The authoring system is designed largely with two functions: cartoon effector and interactive story editor. Cartoon effector automatically converts input images into a cartoon-style image, which consists of image-based cartoon shading and outline drawing steps. Image-based cartoon shading is to receive images of the desired scenes from users, separate brightness information from the color model of the input images, simplify them to a shading range of desired steps, and recreate them as cartoon-style images. Then, the final cartoon style images are created through the outline drawing step in which the outlines of the shaded images are applied through edge detection. Interactive story editor is used to enter text balloons and subtitles in a dialog structure to create one scene of the completed cartoon that delivers a story such as web-toon or comic book. In addition, the cartoon effector, which converts images into cartoon style, is expanded to videos so that it can be applied to videos as well as still images. Finally, various experiments are conducted to verify the possibility of easy and efficient production of cartoons that users want based on images with the proposed IMToon system.

• Journal of KIISE:Computing Practices and Letters
• /
• v.6 no.5
• /
• pp.477-486
• /
• 2000

This paper is for the design and implementation of SGML document presentation system to format SGML document based on DSSSL. ISO/lEC proposed DSSSL as the technological standard for formatting and transforming SGML document. This paper shows the design of this system in accordance with the model defined in DSSSL. This system, which is able to provide Korean, has parsing function of arbitrary DTDs, SGML documents and DSSSL style sheets, and contains a formatter that can manage various specifications, such as graphs, lists, pictures, as well as text. This will satisfy a user's intent for exchange of SGML documents that involve format information between different types, greatly contribute to make up SGML document processing environment.

• Journal of Information Processing Systems
• /
• v.8 no.4
• /
• pp.653-668
• /
• 2012

In this study, an improved HMM based recognition model is proposed for online English and Korean handwritten characters. The pattern elements of the handwriting model are sub character strokes and ligatures. To deal with the problem of handwriting style variations, a modified Hierarchical Clustering approach is introduced to partition different writing styles into several classes. For each of the English letters and each primitive grapheme in Korean characters, one HMM that models the temporal and spatial variability of the handwriting is constructed based on each class. Then the HMMs of Korean graphemes are concatenated to form the Korean character models. The recognition of handwritten characters is implemented by a modified level building algorithm, which incorporates the Korean character combination rules within the efficient network search procedure. Due to the limitation of the HMM based method, a post-processing procedure that takes the global and structural features into account is proposed. Experiments showed that the proposed recognition system achieved a high writer independent recognition rate on unconstrained samples of both English and Korean characters. The comparison with other schemes of HMM-based recognition was also performed to evaluate the system.

• Journal of Information Processing Systems
• /
• v.15 no.6
• /
• pp.1422-1437
• /
• 2019

With the rapid development of the Internet and the Mobile Internet, social communication based on the network has become a life style for many people. WeChat is an online social platform, for about one billion users, therefore, it is meaningful to study the spreading and evolution mechanism of the rumor on the WeChat social circle. The Rumor was injected into the WeChat social circle by certain individuals, and the communication and the evolution occur among the nodes within the circle; after the refuting-rumor-information injected into the circle, subsequently,the density of four types of nodes, including the Susceptible, the Latent, the Infective, and the Recovery changes, which results in evolving the WeChat social circle system. In the study, the evolution characteristics of the four node types are analyzed, through construction of the evolution equation. The evolution process of the rumor injection and the refuting-rumor-information injection is simulated through the structure of the virtual social network, and the evolution laws of the four states are depicted by figures. The significant results from this study suggest that the spreading and evolving of the rumors are closely related to the nodes degree on the WeChat social circle.