• 산업진흥연구
• /
• 제8권4호
• /
• pp.131-137
• /
• 2023

본 연구는 군의 성범죄로 발생하는 문제점을 파악하고자 2019년 2월부터 2022년 5월28일까지 뉴스에서 다룬 군 성범죄 관련 기사를 수집하고 분석하였다. 언론에 보도된 군 성폭력 현황을 파악하고자 뉴스 빅 데이터 전문분석 시스템인 빅카인즈 (BIGKinds)를 활용하여 기사를 수집하였고, Textom 프로그램을 활용해 키워드를 대상으로 시기별 빈도 분석, 워드 클라우드, 의미 연결망 분석 기법을 활용하여 연구를 수행하였다. 데이터 분석 결과, 첫째, 군 내부의 성범죄에 대한 사건 관련 보도는 피해자에게 대중의 관심이 집중된 것을 확인할 수 있었다. 둘째, 성범죄에 대응하는 관련 당국의 미온적 시스템의 문제가 드러났다. 셋째, 성범죄 피해자에 대한 지원 부족이 나타났다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권12호
• /
• pp.3383-3397
• /
• 2023

Scene graphs serve as semantic abstractions of images and play a crucial role in enhancing visual comprehension and reasoning. However, the performance of Scene Graph Generation is often compromised when working with biased data in real-world situations. While many existing systems focus on a single stage of learning for both feature extraction and classification, some employ Class-Balancing strategies, such as Re-weighting, Data Resampling, and Transfer Learning from head to tail. In this paper, we propose a novel approach that decouples the feature extraction and classification phases of the scene graph generation process. For feature extraction, we leverage a transformer-based architecture and design an adaptive calibration function specifically for predicate classification. This function enables us to dynamically adjust the classification scores for each predicate category. Additionally, we introduce a Distribution Alignment technique that effectively balances the class distribution after the feature extraction phase reaches a stable state, thereby facilitating the retraining of the classification head. Importantly, our Distribution Alignment strategy is model-independent and does not require additional supervision, making it applicable to a wide range of SGG models. Using the scene graph diagnostic toolkit on Visual Genome and several popular models, we achieved significant improvements over the previous state-of-the-art methods with our model. Compared to the TDE model, our model improved mR@100 by 70.5% for PredCls, by 84.0% for SGCls, and by 97.6% for SGDet tasks.

• 한국BIM학회 논문집
• /
• 제13권4호
• /
• pp.96-105
• /
• 2023

This study focuses on developing a building object recognition technology for efficient use in the remodeling of buildings constructed without drawings. In the era of the 4th industrial revolution, smart technologies are being developed. This research contributes to the architectural field by introducing a deep learning-based method for automatic object classification and recognition, utilizing point cloud data. We use a TD3D network with voxels, optimizing its performance through adjustments in voxel size and number of blocks. This technology enables the classification of building objects such as walls, floors, and roofs from 3D scanning data, labeling them in polygonal forms to minimize boundary ambiguities. However, challenges in object boundary classifications were observed. The model facilitates the automatic classification of non-building objects, thereby reducing manual effort in data matching processes. It also distinguishes between elements to be demolished or retained during remodeling. The study minimized data set loss space by labeling using the extremities of the x, y, and z coordinates. The research aims to enhance the efficiency of building object classification and improve the quality of architectural plans by reducing manpower and time during remodeling. The study aligns with its goal of developing an efficient classification technology. Future work can extend to creating classified objects using parametric tools with polygon-labeled datasets, offering meaningful numerical analysis for remodeling processes. Continued research in this direction is anticipated to significantly advance the efficiency of building remodeling techniques.

• 대한의용생체공학회:의공학회지
• /
• 제45권2호
• /
• pp.95-100
• /
• 2024

Vitiligo is a condition characterized by the destruction or dysfunction of melanin-producing cells in the skin, resulting in a loss of skin pigmentation. Facial vitiligo, specifically affecting the face, significantly impacts patients' appearance, thereby diminishing their quality of life. Evaluating the efficacy of facial vitiligo treatment typically relies on subjective assessments, such as the Facial Vitiligo Area Scoring Index (F-VASI), which can be time-consuming and subjective due to its reliance on clinical observations like lesion shape and distribution. Various machine learning and deep learning methods have been proposed for segmenting vitiligo areas in facial images, showing promising results. However, these methods often struggle to accurately segment vitiligo lesions irregularly distributed across the face. Therefore, our study introduces a framework aimed at improving the segmentation of vitiligo lesions on the face and providing an evaluation of vitiligo lesions. Our framework for facial vitiligo segmentation and lesion evaluation consists of three main steps. Firstly, we perform face detection to minimize background areas and identify the face area of interest using high-quality ultraviolet photographs. Secondly, we extract facial area masks and vitiligo lesion masks using a semantic segmentation network-based approach with the generated dataset. Thirdly, we automatically calculate the vitiligo area relative to the facial area. We evaluated the performance of facial and vitiligo lesion segmentation using an independent test dataset that was not included in the training and validation, showing excellent results. The framework proposed in this study can serve as a useful tool for evaluating the diagnosis and treatment efficacy of vitiligo.

• 한국컴퓨터정보학회논문지
• /
• 제29권7호
• /
• pp.81-88
• /
• 2024

생성형 AI 시장의 급속한 성장과 국내 기업과 기관의 큰 관심에도 불구하고, 부정확한 정보제공과 정보유출의 우려가 생성형 AI 도입을 저해하는 주된 요인으로 나타났다. 이를 개선하기 위해 본 논문에서는 검색-증강 생성(Retrieval-Augmented Generation, RAG) 구조 기반의 질의응답시스템을 설계·구현하였다. 제안 방법은 한국어 문장 임베딩을 사용해 지식 데이터베이스를 구축하고, 최적화된 검색으로 질문 관련 정보를 찾아 생성형 언어 모델에게 제공된다. 또한, 이용자가 지식 데이터 베이스를 직접 관리하여 변경되는 업무 정보를 효율적으로 업데이트하도록 하고, 시스템이 폐쇄망에서 동작할 수 있도록 설계하여 기업의 기밀 정보의 유출 가능성을 낮추었다. 국내 기업 등 조직에서 생성형 AI를 도입하고 활용하고자 할 때 본 연구가 유용한 참고자료가 되길 기대한다.

• 한국IT서비스학회지
• /
• 제23권3호
• /
• pp.103-118
• /
• 2024

Due to the advancement of virtual reality technology, virtual idols are widely used in industrial and cultural content industries. However, it is difficult to utilize virtual idols' social perceptions because they are not properly understood. Therefore, this paper collected and analyzed YouTube comments to identify differences about social perception through comparative analysis between virtual idols and general idols. The dataset was constructed by crawling comments from music videos with more than 10 million views of virtual idols and more than 10,000 comments. Keyword frequency and TF-IDF values were derived from the collected dataset, and the connection centrality CONCOR cluster was analyzed with a semantic network using the UCINET program. As a result of the analysis, it was found that virtual idols frequently used keywords such as "person," "quality," "character," "reality," "animation," while reactions and perceptions were derived from general idols. Based on the results of this analysis, it was found that while general idols are mainly evaluated with their appearance and cultural factors, social perceptions of virtual idols' values are mixed with evaluations of cultural factors such as "song," "voice," and "choreography," focusing on technical factors such as "people," "quality," "character," and "animation." However, keywords such as "song," "voice," "choreography," and "music" are included in the top 30 like regular idols and appear in the same cluster, suggesting that virtual idols are gradually shifting away from minority tastes to mainstream culture. This study aims to provide academic and practical implications for the future expansion of the industry and cultural content industry of virtual idols by grasping the social perception of virtual idols.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권7호
• /
• pp.1726-1748
• /
• 2024

Trajectory planning is vital for autonomous systems like robotics and UAVs, as it determines optimal, safe paths considering physical limitations, environmental factors, and agent interactions. Recent advancements in trajectory planning and future location prediction stem from rapid progress in machine learning and optimization algorithms. In this paper, we proposed a novel framework for Spatial-temporal transformer-based feed-forward neural networks (STTFFNs). From the traffic flow local area point of view, skip-gram model is trained on trajectory data to generate embeddings that capture the high-level features of different trajectories. These embeddings can then be used as input to a transformer-based trajectory planning model, which can generate trajectories for new objects based on the embeddings of similar trajectories in the training data. In the next step, distant regions, we embedded feedforward network is responsible for generating the distant trajectories by taking as input a set of features that represent the object's current state and historical data. One advantage of using feedforward networks for distant trajectory planning is their ability to capture long-term dependencies in the data. In the final step of forecasting for future locations, the encoder and decoder are crucial parts of the proposed technique. Spatial destinations are encoded utilizing location-based social networks(LBSN) based on visiting semantic locations. The model has been specially trained to forecast future locations using precise longitude and latitude values. Following rigorous testing on two real-world datasets, Porto and Manhattan, it was discovered that the model outperformed a prediction accuracy of 8.7% previous state-of-the-art methods.

• 한국컴퓨터그래픽스학회논문지
• /
• 제30권3호
• /
• pp.109-123
• /
• 2024

딥러닝과 강화학습을 활용한 비전 기반 엔드투엔드 자율주행 시스템 관련 연구가 지속적으로 증가하고 있다. 일반적으로 이러한 시스템은 위치, 속도, 방향, 센서 데이터 등 연속적이고 고차원적인 차량의 상태를 잠재 특징 벡터로 인코딩하고, 이를 차량의 주행 정책으로 디코딩하는 두 단계로 구성된다. 도심 주행과 같이 다양하고 복잡한 환경에서는 Variational Autoencoder(VAE)나 Convolutional Neural Network(CNN)과 같은 네트워크를 이용한 효율적인 상태 표현 방법의 필요성이 더욱 부각된다. 본 논문은 차량의 이미지 상태 표현이 강화학습 성능에 미치는 영향을 분석하였다. CARLA 시뮬레이터 환경에서 실험을 수행하였고, 차량의 전방 카메라 센서로부터 취득한 RGB 이미지 및 Semantic Segmented 이미지를 각각 VAE와 Vision Transformer(ViT) 네트워크로 특징 추출하여 상태 표현 학습에 활용하였다. 이러한 방법론이 강화학습에 미치는 영향을 실험하여, 데이터 유형과 상태 표현 기법이 자율주행의 학습 효율성과 결정 능력 향상에 어떤 역할을 하는지를 실험하였다.

• 복식문화연구
• /
• 제32권4호
• /
• pp.517-546
• /
• 2024

The aim of this study is to investigate the fashion industry's response to climate change and how these discussions unfolded at the 28th Conference of the Parties (COP28) to the United Nations Framework Convention on Climate Change (UNFCCC). Climate change response projects by B Corp-certified fashion companies are examined, focusing on stakeholder efforts and reviewing online media reports. Text data were collected from web documents, interviews, and op-eds relating to COP28 from December 2018 to April 2024 and analyzed using text mining and semantic network analysis to identify critical keywords and contexts. The analysis revealed that the fashion industry is fulfilling its environmental responsibilities through various strategies, prompting changes in consumer behavior by advocating sustainable consumption, including carbon removal, energy transition, and recycling promotion. Stakeholders in online media and those present at COP28 discussed issues relating to climate change in the fashion industry, focusing on environmental protection, energy, greenhouse gas emissions, sustainable material usage, and social responsibility. Key issues at COP28 included policy and regulation, climate change response, energy transition, carbon emissions management, and environmental, social, and governance (ESG) standards. Additionally, by examining the main collections exhibited at the fashion show during COP28, the study analyzed how messages about climate change were conveyed. Fashion companies communicated the industry's response through exhibitions and fashion shows, suggesting a move toward balancing environmental protection and economic growth through the development of sustainable materials, the expansion of recycling and reuse practices, and the modern reinterpretation of cultural heritage.

• 지능정보연구
• /
• 제21권4호
• /
• pp.79-92
• /
• 2015

미래 핵심 가치 기술 발굴 및 탐색을 위해서는 범국가적인 국가R&D정보와 과학기술정보의 연계 융합이 필요하다. 본 논문에서는 국가R&D정보와 과학기술정보를 온톨로지와 토픽모델링을 사용하여 연계 융합하여 지식베이스를 구축한 방법론을 소개하고, 이를 기반으로 한 다차원 연계 지식맵 서비스를 소개한다. 국가R&D정보는 국가R&D과제와 참여인력, 해당 과제에 대한 성과 정보, 논문, 특허, 연구보고서 정보들을 포함한다. 과학기술정보는 논문, 특허, 동향 등의 과학기술연구에 대한 기술 문서를 일컫는다. 본 논문에서는 지식베이스에서의 지식 처리 및 관리의 효율성을 높이기 위해 Lightweight 온톨로지를 사용한다. Lightweight 온톨로지는 국가R&D과제 참여자와 성과정보, 과학기술정보를 과제-성과 관계, 문서-저자 관계, 저자-소속기관 관계 등의 단순한 연관관계를 이용하여 국가R&D정보와 과학기술정보를 융합한다. 이러한 단순한 연관관계만을 이용함으로써 지식 처리의 효율성을 높이고 온톨로지 구축 과정을 자동화한다. 보다 구체적인 Concept 레벨에서의 온톨로지 구축을 위해 토픽모델링을 활용한다. 토픽모델링을 활용하여 국가R&D정보와 과학기술정보 문서들의 토픽 주제어를 추출하고 각 문서 간 연관관계를 추출한다. 일반적인 Concept 레벨에서의 Fully-Specified 온톨로지를 구축하기 위해서는 거의 100% 수동으로 해야 하기 때문에, 많은 시간과 비용이 소모된다. 본 연구에서는 이러한 수동적인 온톨로지 구축이 아닌 자동화된 온톨로지 구축을 위해 토픽모델링을 활용한다. 토픽모델링을 활용하여 온톨로지 구축에 필요한 문서와 토픽 키워드 간의 관계, 문서 간 의미 상 연관관계를 자동으로 추출한다. 마지막으로, 이와 같이 구축된 지식베이스의 트리플(Triple) 정보를 활용하여, 연구자들의 공동저자관계, 문서간의 공통주제어관계 등을 연구자, 주제어, 기관, 저널 등의 다차원 연관관계를 방사형 네트워크 형식을 이용하여 시각화한 지식맵 서비스들을 소개한다.