• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.304-307
• /
• 2021

ML(machine learning) 기술을 활용하여 실용적인 측면에서 일반 사용자들이 바라보고 사용할 수 있도록 다양한 연구 개발이 이루어지고 있다. 특히 최근 개인 사용자의 personal computer와 mobile device의 processing unit의 연산 처리 속도가 두드러지게 빨라지고 있어 ML이 더 생활에 밀접해지고 있는 추세라고 볼 수 있다. 현재 ML시장에서 다양한 솔루션 및 어플리케이션을 제공하는 툴이나 라이브러리가 대거 공개되고 있는데 그 중에서도 Google에서 개발하여 배포한 'Mediapipe'를 사용하였다. Mediapipe는 현재 'android', 'IOS', 'C++', 'Python', 'JS', 'Coral' 등의 환경에서 개발을 지원하고 있으며 더욱 다양한 환경을 지원할 예정이다. 이에 본 팀은 앞서 설명한 Mediapipe 프레임워크를 기반으로 Machine Learning을 사용한 image processing를 통해 일반 사용자들에게 편의성을 제공할 수 있는 알람 프로그램을 연구 및 개발하였다. Mediapipe에서 신체를 landmark로 검출하게 되는데 이를 scikit-learn 머신러닝 라이브러리를 사용하여 특정 자세를 학습시키고 모델화하여 알람 프로그램에 특정 기능에 조건으로 사용될 수 있게 하였다. scikit-learn은 아나콘다 등과 같은 개발환경 패키지에서 간단하게 이용 가능한데 이 아나콘다는 데이터 분석이나 그래프 그리기 등, 파이썬에 자주 사용되는 라이브러리를 포함한 개발환경이라고 할 수 있다. 하여 본 팀은 ML기반의 영상처리 알람 프로그램을 제작하는데에 있어 이러한 사항들을 파이썬 환경에서 기본적으로 포함되어 제공하는 tkinter GUI툴을 사용하고 추가적으로 인텔에서 개발한 실시간 컴퓨터 비전을 목적으로 한 프로그래밍 라이브러리 OpenCV와 여러 항목을 사용하여 환경을 구축할 수 있도록 연구·개발하였다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.07a
• /
• pp.629-632
• /
• 2020

현대사회는 4차 산업혁명 시대가 도래하면서 초연결, 초지능, 초융합 사회로 변화되고 있다. 최근 교육부는 많은 변화가 요구되고 있는 교육분야, 교육정책 방안으로 SW(소프트웨어)교육에 AI(인공지능) 교육까지 추가되야 한다고 제안하고 2024년까지 첨단 기술을 활용한 지능형 과학실을 구축한다고 밝혔다. 이에 본 논문에서는 정부의 교육정책 방안이 원활하게 진행될 수 있고 융합 교육 분야에서 활용될 수 있는 "미래 지능형 과학실 활용을 위한 화학원소기호 이미지 기계학습 AI·SW교육 프로그램"을 제안하고자 한다.

• Journal of Information Processing Systems
• /
• v.19 no.6
• /
• pp.767-777
• /
• 2023

In modern society, user privacy is emerging as an important issue as closed-circuit television (CCTV) systems increase rapidly in various public and private spaces. If CCTV cameras monitor sensitive areas or personal spaces, they can infringe on personal privacy. Someone's behavior patterns, sensitive information, residence, etc. can be exposed, and if the image data collected from CCTV is not properly protected, there can be a risk of data leakage by hackers or illegal accessors. This paper presents an innovative approach to "machine learning based reversible chaotic masking method for user privacy protection in CCTV environment." The proposed method was developed to protect an individual's identity within CCTV images while maintaining the usefulness of the data for surveillance and analysis purposes. This method utilizes a two-step process for user privacy. First, machine learning models are trained to accurately detect and locate human subjects within the CCTV frame. This model is designed to identify individuals accurately and robustly by leveraging state-of-the-art object detection techniques. When an individual is detected, reversible chaos masking technology is applied. This masking technique uses chaos maps to create complex patterns to hide individual facial features and identifiable characteristics. Above all, the generated mask can be reversibly applied and removed, allowing authorized users to access the original unmasking image.

• Proceedings of the IEEK Conference
• /
• summer
• /
• pp.427-434
• /
• 2004

In this paper, we present a overview of content-based image retrieval (CBIR) systems: its results and its problems. We propose our CBIR system currently based on color and texture. From the CBIR systems. we discuss the way to add semantic values in image retrieval systems. There are 3 ways for adding them: concept definition, machine learning and man-machine interaction. Along with this we introduce our preliminary results and discuss them in the goal of reaching semantic retrieval. Different result representation schemes are presented. At last, we present our work to build a complete annotated image database and our image annotaion program.

• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1387-1395
• /
• 2018

Core algorithm of deep learning Convolutional Neural Network(CNN) shows better performance than other machine learning algorithms. However, if there is not sufficient data, CNN can not achieve satisfactory performance even if the classifier is excellent. In this situation, it has been proven that the use of transfer learning can have a great effect. In this paper, we apply two transition learning methods(freezing, retraining) to three CNN models(ResNet-50, Inception-V3, DenseNet-121) and compare and analyze how the classification performance of CNN changes according to the methods. As a result of statistical significance test using various evaluation indicators, ResNet-50, Inception-V3, and DenseNet-121 differed by 1.18 times, 1.09 times, and 1.17 times, respectively. Based on this, we concluded that the retraining method may be more effective than the freezing method in case of transition learning in image classification problem.

• Journal of the Korean Society of Systems Engineering
• /
• v.15 no.2
• /
• pp.72-78
• /
• 2019

Pathology is the motor that drives healthcare to understand diseases. The way pathologists diagnose diseases, which involves manual observation of images under a microscope has been used for the last 150 years, it's time to change. This paper is specifically based on tumor detection using deep learning techniques. Pathologist examine the specimen slides from the specific portion of body (e-g liver, breast, prostate region) and then examine it under the microscope to identify the effected cells among all the normal cells. This process is time consuming and not sufficiently accurate. So, there is a need of a system that can detect tumor automatically in less time. Solution to this problem is computational pathology: an approach to examine tissue data obtained through whole slide imaging using modern image analysis algorithms and to analyze clinically relevant information from these data. Artificial Intelligence models like machine learning and deep learning are used at the molecular levels to generate diagnostic inferences and predictions; and presents this clinically actionable knowledge to pathologist through dynamic and integrated reports. Which enables physicians, laboratory personnel, and other health care system to make the best possible medical decisions. I will discuss the techniques for the automated tumor detection system within the new discipline of computational pathology, which will be useful for the future practice of pathology and, more broadly, medical practice in general.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_1
• /
• pp.507-519
• /
• 2023

Multi-temporal optical images have been utilized for time-series monitoring of croplands. However, the presence of clouds imposes limitations on image availability, often requiring a cloud removal procedure. This study assesses the applicability of various machine learning algorithms for effective cloud removal in optical imagery. We conducted comparative experiments by focusing on two key variables that significantly influence the predictive performance of machine learning algorithms: (1) land-cover types of training data and (2) temporal variability of land-cover types. Three machine learning algorithms, including Gaussian process regression (GPR), support vector machine (SVM), and random forest (RF), were employed for the experiments using simulated cloudy images in paddy fields of Gunsan. GPR and SVM exhibited superior prediction accuracy when the training data had the same land-cover types as the cloud region, and GPR showed the best stability with respect to sampling fluctuations. In addition, RF was the least affected by the land-cover types and temporal variations of training data. These results indicate that GPR is recommended when the land-cover type and spectral characteristics of the training data are the same as those of the cloud region. On the other hand, RF should be applied when it is difficult to obtain training data with the same land-cover types as the cloud region. Therefore, the land-cover types in cloud areas should be taken into account for extracting informative training data along with selecting the optimal machine learning algorithm.

• Journal of Internet of Things and Convergence
• /
• v.5 no.2
• /
• pp.15-20
• /
• 2019

As the machine learning research has developed, the field of translation and image analysis such as optical character recognition has made great progress. However, video translation that combines these two is slower than previous developments. In this paper, we develop an image translator that combines existing OCR technology and translation technology and verify its effectiveness. Before developing, we presented what functions are needed to implement this system and how to implement them, and then tested their performance. With the application program developed through this paper, users can access translation more conveniently, and also can contribute to ensuring the convenience provided in any environment.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.5
• /
• pp.407-414
• /
• 2020

SVM (Support vector machine) analysis was performed after applying a deep learning technique based on an Inception-based model (GoogLeNet). The accuracy of automatic image classification was analyzed using an SVM with multiple virtual grid sizes. Six classes were selected from a standard land cover map. Cars were added as a separate item to increase the classification accuracy of roads. The virtual grid size was 2-5 m for natural areas, 5-10 m for traffic areas, and 10-15 m for building areas, based on the size of items and the resolution of input images. The results demonstrate that automatic classification accuracy can be increased by adopting an integrated approach that utilizes weighted virtual grid sizes for different classes.

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.403-408
• /
• 2018

Generally, a large data set is required for learning of deep learning. However, since it is not easy to create large data sets, there are a lot of techniques that make small data sets larger through data expansion such as rotation, flipping, and filtering. However, these simple techniques have limitation on extendibility because they are difficult to escape from the features already possessed. In order to solve this problem, we propose a method to acquire new image data by using existing data. This is done by retrieving and acquiring similar images using existing image data as a query of the content-based image retrieval (CBIR). Finally, we compare the performance of the base model with the model using CBIR.