• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.287-293
• /
• 2023

With the rapid development of domestic and international over-the-top markets, a large amount of video content is being created. As the volume of video content increases, consumers tend to increasingly check data concerning the videos before watching them. To address this demand, video summaries in the form of plot descriptions, thumbnails, posters, and other formats are provided to consumers. This study proposes an approach that automatically generates posters to effectively convey video content while reducing the cost of video summarization. In the automatic generation of posters, face recognition and clustering are used to gather and classify character data, and keyframes from the video are extracted to learn the overall atmosphere of the video. This study used the facial data of the characters and keyframes as training data and employed technologies such as DreamBooth, a text-to-image generation model, to automatically generate video posters. This process significantly reduces the time and cost of video-poster production.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.8
• /
• pp.2230-2252
• /
• 2024

Advances in deep learning technology have enabled the generation of more realistic deepfakes, which not only endanger individuals' identities but also exploit vulnerabilities in face recognition systems. The majority of existing deepfake detection methods have primarily focused on RGB-based analysis, offering unreliable performance in terms of detection accuracy and time. To address the issue, a grayscale-based deepfake detection method has recently been proposed. This method significantly reduces detection time while providing comparable accuracy to RGB-based methods. However, despite its significant effectiveness, the "key components" that directly affect the performance of grayscale-based deepfake detection have not been systematically analyzed. In this paper, we target three key components: RGB-to-grayscale conversion method, brightness level in grayscale, and resolution level in grayscale. To analyze their impacts on the performance of grayscale-based deepfake detection, we conducted comprehensive evaluations, including component-wise analysis and comparative analysis using real-world datasets. For each key component, we quantitatively analyzed its characteristics' performance and identified differences between them. Moreover, we successfully verified the effectiveness of an optimal combination of the key components by comparing it with existing deepfake detection methods.

• Phonetics and Speech Sciences
• /
• v.15 no.3
• /
• pp.75-82
• /
• 2023

Automatic speech recognition (ASR) has been revolutionized with deep learning-based approaches, among which self-supervised learning methods have proven to be particularly effective. In this study, we aim to enhance the performance of OpenAI's Whisper model, a multilingual ASR system on the Korean language. Whisper was pretrained on a large corpus (around 680,000 hours) of web speech data and has demonstrated strong recognition performance for major languages. However, it faces challenges in recognizing languages such as Korean, which is not major language while training. We address this issue by fine-tuning the Whisper model with an additional dataset comprising about 1,000 hours of Korean speech. We also compare its performance against a Transformer model that was trained from scratch using the same dataset. Our results indicate that fine-tuning the Whisper model significantly improved its Korean speech recognition capabilities in terms of character error rate (CER). Specifically, the performance improved with increasing model size. However, the Whisper model's performance on English deteriorated post fine-tuning, emphasizing the need for further research to develop robust multilingual models. Our study demonstrates the potential of utilizing a fine-tuned Whisper model for Korean ASR applications. Future work will focus on multilingual recognition and optimization for real-time inference.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.15 no.6
• /
• pp.579-589
• /
• 2004

The radar signals are generally very sensitive to relative orientations between radar and target. Thus, the performance of a target recognition system significantly deteriorates as the region of aspect angles becomes broader. To address this difficulty, in this paper, we propose a method based on the multi-aspect information in order to improve the classification capability ever for a wide angular region. First, range profiles are used to extract feature vectors based on the central moments and principal component analysis(PCA). Then, a classifier with the use of multi-aspect information is applied to them, yielding an additional improvement of target recognition capability. There are two different strategies among the classifiers that can fuse the information from multi-aspect radar signals: independent methodology and dependent methodology. In this study, the performances of the two strategies are compared within the frame work of target recognition. The radar cross section(RCS) data of six aircraft models measured at compact range of Pohang University of Science and Technology are used to demonstrate and compare the performances of the two strategies.

• Advances in Computational Design
• /
• v.4 no.2
• /
• pp.155-177
• /
• 2019

Operation and Maintenance (O&M) phase is the main contributor to the total lifecycle cost of a building. Previous studies have described that Building Information Models (BIM), if available with detailed asset information and their properties, can enable rapid troubleshooting and execution of O&M tasks by providing the required information of the facility. Despite the potential benefits, there is still rarely BIM with Mechanical, Electrical and Plumbing (MEP) assets and properties that are available for O&M. BIM is usually not in possession for existing buildings and generating BIM manually is a time-consuming process. Hence, there is a need for an automated approach that can reconstruct the MEP systems in BIM. Previous studies investigated automatic reconstruction of BIM using architectural drawings, structural drawings, or the combination with photos. But most of the previous studies are limited to reconstruct the architectural and structural components. Note that mechanical components in the building typically require more frequent maintenance than architectural or structural components. However, the building mechanical drawings are relatively more complex due to various type of symbols that are used to represent the mechanical systems. In order to address this challenge, this paper proposed a symbol recognition framework that can automatically recognize the different type of symbols in the building mechanical drawings. This study applied vector-based computer vision techniques to recognize the symbols and their properties (e.g., location, type, etc.) in two vector-based input documents: 2D drawings and the symbol description document. The framework not only enables recognizing and locating the mechanical component of interest for BIM reconstruction purpose but opens the possibility of merging the updated information into the current BIM in the future reducing the time of repeated manual creation of BIM after every renovation project.

• Smart Media Journal
• /
• v.11 no.4
• /
• pp.30-37
• /
• 2022

Disease threatens plant growth and recognizing the type of disease is essential to making a remedy. In recent years, deep learning has witnessed a significant improvement for this task, however, a large volume of labeled images is one of the requirements to get decent performance. But annotated images are difficult and expensive to obtain in the agricultural field. Therefore, designing an efficient and effective strategy is one of the challenges in this area with few labeled data. Transfer learning, assuming taking knowledge from a source domain to a target domain, is borrowed to address this issue and observed comparable results. However, current transfer learning strategies can be regarded as a supervised method as it hypothesizes that there are many labeled images in a source domain. In contrast, unsupervised transfer learning, using only images in a source domain, gives more convenience as collecting images is much easier than annotating. In this paper, we leverage unsupervised transfer learning to perform plant disease recognition, by which we achieve a better performance than supervised transfer learning in many cases. Besides, a vision transformer with a bigger model capacity than convolution is utilized to have a better-pretrained feature space. With the vision transformer-based unsupervised transfer learning, we achieve better results than current works in two datasets. Especially, we obtain 97.3% accuracy with only 30 training images for each class in the Plant Village dataset. We hope that our work can encourage the community to pay attention to vision transformer-based unsupervised transfer learning in the agricultural field when with few labeled images.

• The Journal of the Korea Contents Association
• /
• v.12 no.10
• /
• pp.1-10
• /
• 2012

Many statistical methods have been adapted for terminology recognition to improve its accuracy. However, since previous studies have been carried out in a single core or a single machine, they have difficulties in real-time analysing explosively increasing documents. In this study, the task where bottlenecks occur in the process of terminology recognition is classified into linguistic processing in the process of 'candidate terminology extraction' and collection of statistical information in the process of 'terminology weight assignment'. A terminology recognition system is implemented and experimented to address each task by means of the distributed parallel processing-based MapReduce. The experiments were performed in two ways; the first experiment result revealed that distributed parallel processing by means of 12 nodes improves processing speed by 11.27 times as compared to the case of using a single machine and the second experiment was carried out on 1) default environment, 2) multiple reducers, 3) combiner, and 4) the combination of 2)and 3), and the use of 3) showed the best performance. Our terminology recognition system contributes to speed up knowledge extraction of large scale science and technology documents.

• International Journal of Human Ecology
• /
• v.4 no.2
• /
• pp.89-100
• /
• 2003

This paper identifies trends in research methodology due to globalization. Context in both research and in practice and forms the key perspective for modern methodology and theory. Ecological perspectives are a necessary condition for quality global research. Human ecology researchers must advance the role of interdisciplinary and inter-functional perspectives and be open to collaborative relationships. These researchers must work in teams across disciplinary and functional boundaries. The paper discusses directions for research within the context of trends at U.S. federal agencies with applications to globalization and family life. Trends include： (a) use of diverse but rigorous methodologies； (b) recognition of the research-practice-research feedback loop；(c) primacy of context and diverse sampling; and (d) connections of research to problem solving. The terms promoted recently such as ″relationships″, ″diversity″ or ″problem-based″ are ingrained in human ecology. Key aspects for research in the next decade will be： (a) seeking diversity in sampling； (b) seeking colleagues with different perspectives； (c) incorporating meta-analysis into our work； (d) seeking meaningful results； (e) utilizing varieties of research methodologies to address our problems； and (0 understanding that practice must continually change as a function of research.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.11 no.4
• /
• pp.286-292
• /
• 2011

Biometrics systems are used in a wide range of areas, including the area of crime prevention, due to their unique characteristics. However, serious problems can occur if biometric information is disclosed to an unauthorized user. To address these issues, this paper proposes a real valued fuzzy vault method, which adopts a real number error correction code to implement a fuzzy vault scheme for protecting fingerprint temples. The proposed method provides the benefit of allowing the private key value to be changed at any time, unlike biometric template such as a fingerprint, which is not easily renewable even if its security has been breached. The validity of the proposed method is verified for fingerprint verification.

• Journal of Korean Academy of Nursing Administration
• /
• v.23 no.1
• /
• pp.1-7
• /
• 2017

Nursing research in the United States (US) spans several decades. Many of the priorities/trends have stayed through the years. Today, the goal of producing evidence to support nursing care interventions coupled with the drive for Magnet Recognition has encouraged academic nurses (faculty) to work with nurse clinicians to form research teams. Interdisciplinary research teams have also formed to address growing concerns over patient safety and quality care. These issues are not just US issues but global ones. This article addresses US trends with the link to global research trends. The role that organizations such as the International Council of Nurses (ICN), the World Health Organization (WHO), and the Council of International Neonatal Nurses, Inc. (COINN) pay in shaping research agendas and promoting nursing research is highlighted. It emphasizes the key role that nurses, especially nurse leaders/administrators play in changing health outcomes through support of nursing research.