• 방송공학회논문지
• /
• 제23권2호
• /
• pp.186-196
• /
• 2018

본 논문에서는 객체를 추적하기 위해 합성곱 신경망 모델을 이용한 온라인 추적 기법을 제안한다. 오프라인에 모델을 학습시키기 위해서는 많은 수의 훈련 샘플이 필요하다. 이러한 문제를 해결하기 위해, 학습되지 않은 모델을 사용하고, 실험 영상으로부터 직접 훈련 샘플을 수집하여 모델을 갱신한다. 기존의 방법들은 많은 훈련 샘플을 획득하여 모델의 학습에 사용하였지만, 본 논문에서는 적은 수의 훈련 샘플만으로도 객체의 추적이 가능함을 증명한다. 또한 컬러 정보를 활용하여 새로운 손실 함수를 정의하였고 이로부터 잘못 수집된 훈련 샘플로 인해 모델이 잘못된 방향으로 학습되는 문제를 방지한다. 실험을 통해 4가지 비교 방법과 동등하거나 개선된 추적 성능을 보임을 증명하였다.

• 대한기계학회논문집B
• /
• 제30권5호
• /
• pp.489-495
• /
• 2006

Proton exchange membrane (PEM) should be sufficiently hydrated with a careful consideration of heat and water management. Water management has been a critical operation issue for better understanding the operation and optimizing the performance of a PEM fuel cell. The flooding on cathode side resulting from excess water can limit the fuel cell performance. In this study, the visual cell was designed and fabricated fur the visualization of liquid water droplet dynamics related to cathode flooding in flow channels. The experiment was carried out to observe the formation, growth and removal of water droplets using CCD imaging system. Effects of operating conditions such as cell temperature, air flow rate and air relative humidity on cathode flooding characteristics were mainly investigated. Based on this study, we can get the basic insight into flooding phenomena and its two-phase flow nature. It is expected that data obtained can be effectively used fur the setup and validation of two-phase PEM fuel cell models considering cathode flooding.

• 패션비즈니스
• /
• 제4권1호
• /
• pp.1-12
• /
• 2000

The purposes of this thesis are to investigate characteristics, environment of characters through comprehensive approach to analyze its use and strategies in apparel brands. As a method to accomplish this research, fashion related articles, documents and magazines are used along with marketing references forecasting 21st century market changes. The character industry has diverse application to different medias and also intimacy and absorbtion beyond sex, age, generation, nationality. The cultural background of character fashion is based on casual clothing caused by wide spread pursuit of sports and leisure culture, indivisualization and diversification of clothing, customer-made marketing atmosphere and tendency of pursuit of fun and humor. In case of domestic young casual market, own characters are developed for creating differentiated it's own brand images. Characters are applied as an design details or cyber fashion models standing for the image of it's own brand and take a part in other events and visual parts as well. Characters not only limited to clothing items but also further usages of characters extended to stickers and other related stationary goods are necessary. A 'Character multi shop' can be one of good methods to maximize synergy effect.

• 대한원격탐사학회지
• /
• 제19권3호
• /
• pp.255-261
• /
• 2003

Recently, the enormous increase in the volume of remotely sensed data is being acquired by an ever-growing number of earth observation satellites. The combining of diversely sourced imagery together is an important requirement in many applications such as data fusion, city modeling and object recognition. Aerial triangulation is a procedure to reconstruct object space from imagery. However, since the different kinds of imagery have their own sensor model, characteristics, and resolution, the previous approach in aerial triangulation (or georeferencing) is purformed on a sensor model separately. This study evaluated the advantages of aerial triangulation of large number of images from multi-sensors simultaneously. The incorporated multi-sensors are frame, push broom, and whisky broom cameras. The limits and problems of push-broom or whisky broom sensor models can be compensated by combined triangulation with other sensors The reconstructed object space from multi-sensor triangulation is more accurate than that from a single model. Experiments conducted in this study show the more accurately reconstructed object space from multi-sensor triangulation.

• 대한건축학회연합논문집
• /
• 제20권6호
• /
• pp.71-80
• /
• 2018

The path and visibility of nurses in general hospital wards have been treated as architectural planning factors. However, the analysis approach of existing studies shows limitations that only fixed physical elements are considered without considering the behavior of users using space. Consider factors for analysis of ward and models based on this study model. Select a case hospital to apply the analysis technique and conduct the nurse questionnaire and route survey of the hospital. Establish a framework for analysis model applied with the path of nurse movements. The analysis model applies to the case hospital. The analysis results are aggregated to derive design suggestion for reference to the spatial improvement of the ward. Visible visibility to observe the bedside in the nursing station, visibility to observe the beds in the nurse's path, and visibility to observe patients moving in the nurse's path were derived from visual access frequency and exposure frequency. The survey of nurses' movements at the site allowed the nurses to calculate the distance required to move. Reflecting the path of nurse movement, a model was presented for a comprehensive analysis of nursing distance and nursing visibility, which could lead to improvement in the observation and visibility of nurses and the layout of patient rooms or day rooms.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권8호
• /
• pp.3942-3961
• /
• 2019

Recently, the models of deep super-resolution networks can successfully learn the non-linear mapping from the low-resolution inputs to high-resolution outputs. However, for large scaling factors, this approach has difficulties in learning the relation of low-resolution to high-resolution images, which lead to the poor restoration. In this paper, we propose Stage Generative Adversarial Networks (Stage-GAN) with semantic maps for image super-resolution (SR) in large scaling factors. We decompose the task of image super-resolution into a novel semantic map based reconstruction and refinement process. In the initial stage, the semantic maps based on the given low-resolution images can be generated by Stage-0 GAN. In the next stage, the generated semantic maps from Stage-0 and corresponding low-resolution images can be used to yield high-resolution images by Stage-1 GAN. In order to remove the reconstruction artifacts and blurs for high-resolution images, Stage-2 GAN based post-processing module is proposed in the last stage, which can reconstruct high-resolution images with photo-realistic details. Extensive experiments and comparisons with other SR methods demonstrate that our proposed method can restore photo-realistic images with visual improvements. For scale factor ${\times}8$, our method performs favorably against other methods in terms of gradients similarity.

• 전기전자학회논문지
• /
• 제22권4호
• /
• pp.1044-1049
• /
• 2018

CNN(Convolutional Neural Networks)은 동물의 시각정보처리과정을 모델링한 신경망으로 다양한 분야에서 좋은 성능을 보여주고 있다. 본 논문에서는 CNN을 사용하여 능동소나 신호의 스펙트로그램을 분석하고, 표적과 비표적을 식별하는 연구를 수행하였다. 데이터를 표적이 포함된 비율에 따라 8클래스로 구분하고, CNN의 학습에 사용하였다. 신호의 스펙트로그램을 프레임별로 나누어 입력으로 사용한 결과, 표적신호의 위치에서만 표적신호에 해당하는 7개 클래스의 식별 결과가 순차적으로 나타나는 특성을 사용하여 표적과 비표적을 식별해낼 수 있었다.

• 한국멀티미디어학회논문지
• /
• 제21권12호
• /
• pp.1417-1424
• /
• 2018

In order to improve the performance of image classifications using Convolutional Neural Networks (CNN), applying a category hierarchy to the classification can be a useful idea. However, the visual separation of object categories is very different according to the upper and lower category levels and highly uneven in image classifications. Therefore, it is doubtable whether the use of category hierarchies for classification is effective in CNN. In this paper, we have clarified whether the image classification using category hierarchies improves classification performance, and found at which level of hierarchy classification is more effective. For experiments we divided the image classification task according to the upper and lower category levels and assigned image data to each CNN model. We identified and compared the results of three classification models and analyzed them. Through the experiments, we could confirm that classification effectiveness was not improved by reduction of number of categories in a classification model. And we found that only with the re-training method in the last network layer, the performance of lower category classification was not improved although that of higher category classification was improved.

• 국제초고층학회논문집
• /
• 제9권4호
• /
• pp.351-360
• /
• 2020

Monitoring civil structures periodically is necessary for ensuring the fitness of the structures. Cracks on inner and outer surfaces of the building plays a vital role in indicating the health of the building. Conventionally, human visual inspection techniques were carried up to human reachable altitudes. Monitoring of high rise infrastructures cannot be done using this primitive method. Also, there is a necessity for more accurate prediction of cracks on building surfaces for ensuring the health and safety of the building. The proposed research focused on developing an efficient crack classification model using Transfer Learning enabled EfficientNet (TL-EN) architecture. Though many other pre-trained models were available for crack classification, they rely on more number of training parameters for better accuracy. The TL-EN model attained an accuracy of 0.99 with less number of parameters on large dataset. A bench marked METU dataset with 40000 images were used to test and validate the proposed model. The surfaces of high rise buildings were investigated using vision enabled Unmanned Arial Vehicles (UAV). These UAV is fabricated with TL-EN model schema for capturing and analyzing the real time streaming video of building surfaces.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권3호
• /
• pp.877-893
• /
• 2022

With the development of deep learning, face inpainting has been significantly enhanced in the past few years. Although image inpainting framework integrated with generative adversarial network or attention mechanism enhanced the semantic understanding among facial components, the issues of reconstruction on corrupted regions are still worthy to explore, such as blurred edge structure, excessive smoothness, unreasonable semantic understanding and visual artifacts, etc. To address these issues, we propose a Learnable Structure Knowledge of Fusion Network (LSK-FNet), which learns a prior knowledge by edge generation network for image inpainting. The architecture involves two steps: Firstly, structure information obtained by edge generation network is used as the prior knowledge for face inpainting network. Secondly, both the generated prior knowledge and the incomplete image are fed into the face inpainting network together to get the fusion information. To improve the accuracy of inpainting, both of gated convolution and region normalization are applied in our proposed model. We evaluate our LSK-FNet qualitatively and quantitatively on the CelebA-HQ dataset. The experimental results demonstrate that the edge structure and details of facial images can be improved by using LSK-FNet. Our model surpasses the compared models on L1, PSNR and SSIM metrics. When the masked region is less than 20%, L1 loss reduce by more than 4.3%.