• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권8호
• /
• pp.2881-2894
• /
• 2014

In complicated environment, context information plays an important role in image segmentation/labeling. The recently proposed auto-context algorithm is one of the effective context-based methods. However, the standard auto-context approach samples the context locations utilizing a fixed radius sequence, which is sensitive to large scale-change of objects. In this paper, we present a scale invariant auto-context (SIAC) algorithm which is an improved version of the auto-context algorithm. In order to achieve scale-invariance, we try to approximate the optimal scale for the image in an iterative way and adopt the corresponding optimal radius sequence for context location sampling, both in training and testing. In each iteration of the proposed SIAC algorithm, we use the current classification map to estimate the image scale, and the corresponding radius sequence is then used for choosing context locations. The algorithm iteratively updates the classification maps, as well as the image scales, until convergence. We demonstrate the SIAC algorithm on several image segmentation/labeling tasks. The results demonstrate improvement over the standard auto-context algorithm when large scale-change of objects exists.

• 한국멀티미디어학회논문지
• /
• 제22권11호
• /
• pp.1251-1258
• /
• 2019

Intelligent tutoring system enables users to effectively learn by utilizing various artificial intelligence techniques. For instance, it can recommend a proper curriculum or learning method to individual users based on their learning history. To do this effectively, user's characteristics need to be analyzed and classified based on various aspects such as interest, learning ability, and personality. Even though data labeled by the characteristics are required for more accurate classification, it is not easy to acquire enough amount of labeled data due to the labeling cost. On the other hand, unlabeled data should not need labeling process to make a large number of unlabeled data be collected and utilized. In this paper, we propose a semi-supervised learning method based on feedback variational auto-encoder(FVAE), which uses both labeled data and unlabeled data. FVAE is a variation of variational auto-encoder(VAE), where a multi-layer perceptron is added for giving feedback. Using unlabeled data, we train FVAE and fetch the encoder of FVAE. And then, we extract features from labeled data by using the encoder and train classifiers with the extracted features. In the experiments, we proved that FVAE-based semi-supervised learning was superior to VAE-based method in terms with accuracy and F1 score.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2022년도 추계학술대회
• /
• pp.222-224
• /
• 2022

객체 분할 분야의 딥러닝 기반 컴퓨터 비전 응용들은 성능을 향상하기 위하여 STOA 기법들이 사전학습하여 배포한 하이퍼파라미터와 모델을 통해 학습하는 전이학습 방법을 사용한다. 이 과정에서 사용되는 커스텀 데이터 셋들은 Ground Truth 정보를 생성하기 위한 라벨링 작업에서 시간이나 라벨러등의 많은 자원을 필요로 한다. 본 고에서는 딥러닝 신경망에서 사용되는 커스텀 데이터 셋 구축을 위하여 시간이나 라벨러등의 자원을 적게 사용할 수 있도록 객체 분할 기법을 활용한 자동 라벨링 구축 방법을 제시한다.

• 전기전자학회논문지
• /
• 제26권3호
• /
• pp.341-348
• /
• 2022

Wi-Fi가 거의 모든 곳에서 사용이 가능한 환경이 도래하면서 Wi-Fi 기반의 센싱 시스템의 활용가능성에 대한 학계의 주목과 함께 활발한 연구가 진행되고 있다. 최근에는 채널 상태 정보(CSI)를 활용한 딥러닝 기술의 비약적 발달로 높은 감지 성능을 달성하고 있다. 하지만, 새로운 대상 도메인에 적용하기 위해서는 명시적인 데이터 수집 및 모델 재학습 과정의 값비싼 적응 노력 없이는 여전히 실질적으로는 사용하기가 어렵다. 본 연구에서는 딥러닝 기반의 Wi-Fi 센싱 시스템을 위한 훈련데이터 수집 및 레이블링을 자동으로 진행하는 CSI 자동 레이블링 시스템(CALS)를 제안한다. 제안 시스템은 CSI 데이터 수집 과정에서 컴퓨터 비전 기술을 함께 활용하여, 지도학습용으로 수집된 CSI 데이터에 대한 레이블링을 자동으로 수행토록 하였다. CALS의 효율성을 보이기 위해 라즈베리파이를 이용하여 프로토타입 시스템을 구현하고, 실내 환경에서의 사람 존재 감지를 수행하는 3가지 모델에 대해 학습과 평가를 진행하였다. 자동 수집된 데이터를 진행하여 학습을 활용하는 방식으로 실시간 데이터에 대해 평가를 진행했을 때 90% 이상의 높은 정확도를 달성하였다.

• 한국컴퓨터정보학회논문지
• /
• 제24권9호
• /
• pp.21-27
• /
• 2019

In this paper, we propose a deep auto-encoder-based pipe leak detection (PLD) technique from time-series acoustic data collected by microphone sensor nodes. The key idea of the proposed technique is to learn representative features of the leak-free state using leak-free time-series acoustic data and the deep auto-encoder. The proposed technique can be used to create a PLD model that detects leaks in the pipeline in an unsupervised learning manner. This means that we only use leak-free data without labeling while training the deep auto-encoder. In addition, when compared to the previous supervised learning-based PLD method that uses image features, this technique does not require complex preprocessing of time-series acoustic data owing to the unsupervised feature extraction scheme. The experimental results show that the proposed PLD method using the deep auto-encoder can provide reliable PLD accuracy even considering unsupervised learning-based feature extraction.

• 한국지능시스템학회논문지
• /
• 제25권6호
• /
• pp.585-591
• /
• 2015

구어 연구를 위한 전사 과정에서 문자로 표현된 발화를 녹음 음성에 연결해주는 작업을 레이블링이라고 한다. 기존 레이블링 도구들은 대부분 수동으로 작업이 이루어진다. 제안하는 반자동 레이블링은 자동화 모듈과 수동 조정 모듈로 구성된다. 자동화 모듈은 G.Saha 알고리즘을 활용하여 음성구간을 추출하고, 기구축된 발화텍스트의 발화 수와 발화의 길이 정보를 이용하여 발화구간을 예측한다. 본 논문에서는 기존 수동 도구의 정확성을 유지하기 위하여 자동 레이블링된 발화구간을 보정하기 위한 수동 조정 사용자 인터페이스를 제공한다. 제안하는 반자동 레이블링 알고리즘으로 구현한 도구는 기존 수동 레이블링 도구와 비교하여 작업 속도가 평균 27% 향상되었다.

• Nutrition Research and Practice
• /
• 제9권1호
• /
• pp.92-98
• /
• 2015

BACKGROUND/OBJECTIVES: The prevalence of food allergies in Korean children aged 6 to 12 years increased from 10.9% in 1995 to 12.6% in 2012 according to nationwide population studies. Treatment for food allergies is avoidance of allergenic-related foods and epinephrine auto-injector (EPI) for accidental allergic reactions. This study compared knowledge and perception of food allergy labeling and dietary practices of students. SUBJECTS/METHODS: The study was conducted with the fourth to sixth grade students from an elementary school in Yongin. A total of 437 response rate (95%) questionnaires were collected and statistically analyzed. RESULTS: The prevalence of food allergy among respondents was 19.7%, and the most common food allergy-related symptoms were urticaria, followed by itching, vomiting and nausea. Food allergens, other than 12 statutory food allergens, included cheese, cucumber, kiwi, melon, clam, green tea, walnut, grape, apricot and pineapple. Children with and without food allergy experience had a similar level of knowledge on food allergies. Children with food allergy experience thought that food allergy-related labeling on school menus was not clear or informative. CONCLUSION: To understand food allergies and prevent allergic reactions to school foodservice among children, schools must provide more concrete and customized food allergy education.

• 한국조경학회지
• /
• 제23권4호
• /
• pp.20-27
• /
• 1996

The main purpose of this research is to develop a program supporting landscape planting design on AutoCAD basis using AutoLISP and DCL language. Current CAD use in landscape architecture field is mainly focused on customizing plant symbols for supporting two dimensional drafting rather than three dimensional consideration. This program is composed of eight module a such as PLANT module for inserting plant symbols, LABEL module for labeling task, SIMULATION module for simulating plant growth and seasonal color variation, TABLE module for generating plant table automatically, BUILDING module, BLOCK module, UTILITY module for deleting, transforming, shading symbols and DB MANAGER module for manipulating data. Design automation ability using automatic object recognition technique in this program allows AutoCAD to be used as a design tool in addition to its main role as a drafting tool through supporting landscape designers to generate many alternatives in the early phase of design.

• Maxillofacial Plastic and Reconstructive Surgery
• /
• 제31권3호
• /
• pp.207-215
• /
• 2009

Purpose: The aim of the present study is to evaluate the long term bone healing after horizontal ridge augmentation using auto block bone graft for implant installation timing. Materials and Methods: Five Beagle dogs(which were 14 months old and weighted approximately 10kg). In surgery 1(extraction & bone defect), premolars(P2, P3,P4) were extracted and the buccal bone plate was removed to create a horizontally defected ridge. After three months healing, in surgery 2(ridge augmentation). Auto block bone grafts from the mandibular ramus were used in filling the bone defects were fixed with stabilizing screws. The following fluorochrome labels were given intravenously to the beagle dogs: oxytetracycline 1week after the surgery, alizarin red 4 weeks after the surgery, calcein blue 8 weeks after the surgery. The tissue samples were obtained from the sacrificed dogs of 1, 4, 8, 12, 16 weeks after the surgery. Non-decalcified sections were prepared by resin embedding and microsection to find thickness of $10{\mu}m$ for the histologic examination and analysis. Results: 1. We could achieve the successful reconstruction of the horizontal bone defect by auto block bone graft. The grafted bone block remained stable morohologically after 16 weeks of the surgery. 2. In the histologic view. We observed osteoid tissue from the sample $4^{th}$ week sample and active capillary reconstruction in the grafted bone from the $12^{th}$ week sample. Healing procedures of auto bone grafts were compared to that of the host bone. 3. Bone mineralization could be detected from the $8^{th}$ week sample. 4. Fluorochrome labeling showed active bony changes and formation at the interface of the host bone and the block graft mainly. Bony activation in the grafted bone could be seen from the $4^{th}$ week samples. Conclusions: Active bone formation and remodeling between the grafted bone and host bone can be seen through the revascularization. After the perfect adhesion to host bone, Timing of successful implant installation can be detected through the ideal ridge formation by horizontal ridge augmentation.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1996년도 하계학술대회 논문집 B
• /
• pp.1266-1268
• /
• 1996

In this study an automatic roiling-coli labeling system using robot vision system and peripheral mechanism is proposed and implemented, which instead of the manual labor to attach labels Rolling-coils in a steel miil. The binary image process for the image processing is performed with the threshold, and the contour line is converted to the binary gradient which detects the discontinuous variation of brightness of rolling-coils. The moment invariants algorithm proposed by Hu is used to make it easy to recognize even when the position of the center are different from the trained data. The position error compensation algorithm of six degrees of freedom industrial robot manipulator is also developed and the data of the position of the center rolling-coils, which is obtained by floor mount camera, are transfered by asynchronous communication method. Therefore even if the position of center is changed, robot moves to the position of center and performs the labeling work successfully. Therefore, this system can be improved the safety and efficiency.