• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.01a
• /
• pp.27-29
• /
• 2022

사상의학은 외형과 병증 등을 바탕으로 체질을 감별하고 이를 진단에 활용하는 한국의 고유 체질 의학이다. 체형은 체질 변증의 중요한 단서로, 계측정보를 사용한 체질별 도식화 및 감별을 위한 기존 연구가 있었으나, 한정된 샘플수와 연구 간의 이질성으로 대규모 집단 분석 결과가 도출되기 어려우며, 실측 및 라벨링 데이터가 필수적이라는 한계가 있다. 본 연구는 한국인 체형 빅데이터를 사용하여, 영상 정보만으로 체질 감별에 필요한 체형 요소를 추출하고, 이를 기존 문헌에서 제시한 체질 감별 공식에 적용하여 사상체질을 자동 감별하는 모델을 제안한다.

• Annual Conference of KIPS
• /
• 2023.05a
• /
• pp.485-487
• /
• 2023

컴퓨터 비전의 하위 태스크(Task)인 의미론적 분할(Semantic Segmentation)은 자율주행, 해상에서 선박찾기 등 다양한 분야에서 연구되고 있다. 기존 FCN(Fully Conovlutional Networks) 기반 의미론적 분할 모델은 다운샘플링(Dowsnsampling)과정에서 공간정보의 손실이 발생하여 정확도가 하락했다. 본 논문에서는 공간정보 손실을 완화하고자 PSA(Polarized Self-attention)의 공간정보 강조 모듈을 HRNet(High-resolution Networks)의 합성곱 블록 사이에 추가한다. 실험결과 파라미터는 3.1M, GFLOPs는 3.2G 증가했으나 mIoU는 0.26% 증가했다. 공간정보가 의미론적 분할 정확도에 영향이 미치는 것을 확인했다.

• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.914-915
• /
• 2023

COVID-19 대유행 및 이태원 압사 사고로 인해 안전 관점의 인구 밀집에 관심이 높아졌으며, 기존 CCTV를 통한 단순 관찰방식을 넘어 유동 인구의 흐름까지 예측한 인구 밀집도 파악이 필요하게 되었다. 본 논문에서는 기존 관찰방식 공공데이터 CCTV에 컴퓨터 비전(CV) 및 다중 객체 추적(MOT) 기술을 추가 적용하여 사용자 중심(시각, 장소)의 유동 인구수와 인구 밀집 지역을 파악할 수 있는 모델을 제안하고 구현하였다. 이 모델을 적용함으로써 시민들은 안전한 환경에서 인구 밀집에 관련된 사고로부터 보호받을 수 있을 것으로 기대한다.

• Annual Conference of KIPS
• /
• 2007.05a
• /
• pp.267-270
• /
• 2007

효과적인 소프트웨어 프로세스 개선을 위해 SPICE와 CMMI 프로세스 심사 표준을 도입하려는 노력을 하고 있다. 이러한 표준을 통해 효과적인 개선하기 위해서는 개선점과 위험을 식별하고 이들 이슈들을 개발환경에 적용시켜서 조직의 비전에 대응한 작업성능을 높여야한다. 지속적인 개선을 필요로 하는 조직은 현재의 작업성능을 측정하고 이를 개선하기 위한 개선점을 찾아내는 능력과 경험을 축적하여 체계적으로 관리하는 것이 중요하다. 하지만 기존의 SPI 모델들은 무엇을 수행해야 하는지에 대한 지침은 제공하고 있지만, 정량적인 작업성능 측정 및 특정 환경의 소프트웨어 개발 조직의 SPI를 위해 필요한 구체적인 지침을 제시하고 있지는 않다. 따라서, 본 논문에서는 정량적인 SPI룰 위해 프로세스 측정 메트릭 정의와 심사 경험이 분석되어 활용될 수 있는 PCM(Process Capability Metric) Experience Factory 모델을 제안한다.

• Journal of the Korea Computer Graphics Society
• /
• v.29 no.3
• /
• pp.79-84
• /
• 2023

Image generation models have been applied in various fields to overcome data sparsity, time and cost issues. However, it has limitations in generating images from regular pattern images and detecting defects in such data. In this paper, we verified the feasibility of the image generation model to generate pattern images and applied it to data augmentation for defect detection of OLED panels. The data required to train an OLED defect detection model is difficult to obtain due to the high cost of OLED panels. Therefore, even if the data set is obtained, it is necessary to define and classify various defect types. This paper introduces an OLED panel defect data acquisition system that acquires a hypothetical data set and augments the data with an image generation model. In addition, the difficulty of generating pattern images in the diffusion model is identified and a possibility is proposed, and the limitations of data augmentation and defect detection data augmentation using the image generation model are improved.

• Smart Media Journal
• /
• v.1 no.2
• /
• pp.40-46
• /
• 2012

In this paper, we describe interactive contents which is used the result of the inputted interface recognizing vision-based body gesture. Because the content uses the imp which is the common culture as the subject in Asia, we can enjoy it with culture familiarity. And also since the player can use their own gesture to fight with the imp in the game, they are naturally absorbed in the game. And the users can choose the multiple endings of the contents in the end of the scenario. In the part of the gesture recognition, KINECT is used to obtain the three-dimensional coordinates of each joint of the limb to capture the static pose of the actions. The vision-based 3D human pose recognition technology is used to method for convey human gesture in HCI(Human-Computer Interaction). 2D pose model based recognition method recognizes simple 2D human pose in particular environment On the other hand, 3D pose model which describes 3D human body skeletal structure can recognize more complex 3D pose than 2D pose model in because it can use joint angle and shape information of body part Because gestures can be presented through sequential static poses, we recognize the gestures which are configured poses by using HMM In this paper, we describe the interactive content which is used as input interface by using gesture recognition result. So, we can control the contents using only user's gestures naturally. And we intended to improve the immersion and the interest by using the imp who is used real-time interaction with user.

• Journal of Bio-Environment Control
• /
• v.32 no.4
• /
• pp.456-465
• /
• 2023

Melon fruits exhibit a wide range of morphological variations in fruit shape, sugar content, net quality, diameter and weight, which are largely dependent on the variety. These characteristics significantly affect marketability. For netted varieties, the uniformity and pattern of the net serve as key factors in determining the external quality of the melon and act as indicators of its internal quality. In this study, we evaluated the effect of fruit morphology and growth on netting by analyzing the changes in melon fruit quality under LED light treatment and monitoring fruit growth. Computer vision analysis was used for quantitative evaluation of fruit net quality, and a three-variable logistic model was applied to simulate fruit growth. The results showed that melons grown under LED conditions exhibited more uniform fruit shape and improvements in both net quality and sugar content compared to the control group. The results of the logistic model showed minimal error values and consistent curve slopes across treatments, confirming its ability to accurately predict fruit growth patterns under varying light conditions. This study provides an understanding of the effects of fruit shape and growth on net quality.

• The Journal of Bigdata
• /
• v.7 no.2
• /
• pp.11-29
• /
• 2022

Pre-trained language models are the most important and widely used tools in natural language processing tasks. Since those have been pre-trained for a large amount of corpus, high performance can be expected even with fine-tuning learning using a small number of data. Since the elements necessary for implementation, such as a pre-trained tokenizer and a deep learning model including pre-trained weights, are distributed together, the cost and period of natural language processing has been greatly reduced. Transformer variants are the most representative pre-trained language models that provide these advantages. Those are being actively used in other fields such as computer vision and audio applications. In order to make it easier for researchers to understand the pre-trained language model and apply it to natural language processing tasks, this paper describes the definition of the language model and the pre-learning language model, and discusses the development process of the pre-trained language model and especially representative Transformer variants.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.3
• /
• pp.513-520
• /
• 2020

Recently, the development of a vision inspection system using machine learning has become more active. This study seeks to develop a defect inspection model using machine learning. Defect detection problems for images correspond to classification problems, which are the method of supervised learning in machine learning. In this study, defect detection models are developed based on algorithms that automatically extract features and algorithms that do not extract features. One-dimensional CNN and two-dimensional CNN are used as algorithms for automatic extraction of features, and MLP and SVM are used as algorithms for non-extracting features. A defect detection model is developed based on four models and their accuracy and AUC compare based on AUC. Although image classification is common in the development of models using CNN, high accuracy and AUC is achieved when developing SVM models by converting pixels from images into RGB values in this study.

• Journal of Convergence for Information Technology
• /
• v.7 no.3
• /
• pp.73-81
• /
• 2017

The generalized Hough transform (GHough) is a useful technique for detecting and locating 2-D model. However, GHough requires a 4-D parameter array and a large amount of time to detect objects of unknown scale and orientation because it enumerates all possible parameter values into a 4-D parameter space. Several n-to-1 mapping algorithms were proposed to reduce the parameter space from 4-D to 2-D. However, these algorithms are very likely to fail due to the random votes cast into the 2-D parameter space. This paper proposes to use internal gradient information in addition to the model boundary points to reduce the number of random votes cast into 2-D parameter space. Experimental result shows that our proposed method can reduce both the number of random votes cast into the parameter space and the execution time effectively.