소셜 네트웍 환경에서의 얼굴 어노테이션

최권택(Kwontaeg Choi),변혜란(Hyeran Byun) 한국정보과학회 2008 한국정보과학회 학술발표논문집 Vol.35 No.2

단어 임베딩을 이용한 사진 유사도 평가 및 가상현실기반 인지재활 시스템

최권택(KwonTaeg Choi) 한국디지털콘텐츠학회 2019 한국디지털콘텐츠학회논문지 Vol.20 No.11

As the dementia elderly increase due to aging, socio-economic costs have increased greatly and research is being done to solve this problem.. This paper proposes a virtual reality-based cognitive rehabilitation system using image recognition technology and clustering algorithm by collecting pictures reflecting memories and memories of subjects. It is distinguished from existing methods in that it is a converged content that reproduces individual experiences and memories in the virtual reality space by using existing objective photos or pictures. We evaluated the effectiveness of the proposed clustering algorithm using 2022 photographs as an experimental data set. By expressing images with semantic-based word vectors, we confirmed that contextually similar photographs were clustered. We also found that the method of selecting photos to maintain the minimum distance from the Gaussian mixture model is very efficient for selecting contextually separated photos. In addition, a virtual reality-based cognitive rehabilitation system using the proposed algorithm is presented.

IoT 디바이스에서 다차원 디지털 신호 처리를 위한 신경망 최적화

최권택(KwonTaeg Choi) 한국디지털콘텐츠학회 2017 한국디지털콘텐츠학회논문지 Vol.18 No.6

Deep learning method, which is one of the most famous machine learning algorithms, has proven its applicability in various applications and is widely used in digital signal processing. However, it is difficult to apply deep learning technology to IoT devices with limited CPU performance and memory capacity, because a large number of training samples requires a lot of memory and computation time. In particular, if the Arduino with a very small memory capacity of 2K to 8K, is used, there are many limitations in implementing the algorithm. In this paper, we propose a method to optimize the ELM algorithm, which is proved to be accurate and efficient in various fields, on Arduino board. Experiments have shown that multi-class learning is possible up to 15-dimensional data on Arduino UNO with memory capacity of 2KB and possible up to 42-dimensional data on Arduino MEGA with memory capacity of 8KB. To evaluate the experiment, we proved the effectiveness of the proposed algorithm using the data sets generated using gaussian mixture modeling and the public UCI data sets.

개방형 웹 서비스를 위한 증가적 얼굴 어노테이션

최권택(Kwontaeg Choi),변혜란(Hyeran Byun) 한국정보과학회 2009 정보과학회논문지 : 소프트웨어 및 응용 Vol.36 No.8

최근 Flickr, Facebook, Cyworld 처럼 사진 공유를 기반으로 하는 소셜 웹 서비스의 성공과 발달로 얼굴 검출/인식 기술을 이러한 서비스에 접목하려는 다양한 시도가 진행되고 있다. 그러나 인식률 향상에만 초점을 맞춘 기존의 일관처리 기반의 연구들은 수백만의 이용자가 수시로 접근하는 서비스에 적용하기 어렵다. 본 논문에서는 시간에 따라 증가하는 거대한 얼굴 영상 데이터베이스를 효과적으로 분류하기 위해 랜덤 사상(Random Projectio, RP) 비선형 회귀(Non-linear Regression) 그리고 REST(REpresentational State Transfer) 규약을 사용해 새로운 증가적 얼굴 어노테이션 방법을 제안하고자 한다. 다양한 비교실험 결과에서 제안된 방법은 향상된 인식률과 낮은 계산 복잡도 기록했다. 따라서 제안된 방법은 대규모 웹서비스에 적합한 얼굴 어노테이션 알고리즘이다. Recently, photo sharing and publishing based Social Network Sites(SNSs) are increasingly attracting the attention of academic and industry researches. Unlike the face recognition environment addressed by existing works, face annotation problem under SNSs is differentiated in terms of daily updated images database, a limited number of training set and millions of users. Thus, conventional approach may not deal with these problems. In this paper, we proposed a face annotation method for sharing and publishing photographs that contain faces under a social network service using random projection, non-linear regression and representational state transfer. Our experiments on several databases show that the proposed method records an almost constant execution time with comparable accuracy of the PCA-SVM classifier.

소셜 네트웍 환경에서의 얼굴 주석 시스템

최권택(Kwontaeg Choi),변혜란(Hyeran Byun) 한국정보과학회 2009 정보과학회 컴퓨팅의 실제 논문지 Vol.15 No.8

최근 사진 공유 기반의 소셜 네트웍 서비스의 발달로 수백만 명의 사람들이 인터넷 공간에서 온라인 커뮤니티 활동에 참여하고 있다. 본 논문에서는 이러한 소셜네트웍 서비스 환경에서 얼굴 사진에 주석 정보를 부여하고 이를 검색할 수 있는 효과적인 방법론을 제안한다. 지속적으로 이용자와 이미지가 증가하는 방대한 데이터베이스를 취급해야하기 때문에 인식률 뿐만 아니라 계산 복잡도가 매우 낮아야 한다. 본 논문에 이러한 문제를 해결하기 위해 온라인 학습과 사회적 관계를 이용한 다중 분류기를 제안한다. 실험결과를 통해 제안된 방법은 보편적으로 사용되는 서포트 백터 머신과 비교해 향상된 인식률과 낮은 계산 복잡도를 보여줌으로써 사용자의 주석 횟수를 줄이고, 사용자에게 빠른 응답을 할 수 있음을 보여준다. Recently, photo sharing and publishing based Social Network Sites(SNSs) are increasingly attracting the attention of academic and industry researches. Millions of users have integrated these sites into their daily practices to communicate with online people. In this paper, we propose an efficient face annotation and retrieval system under SNS. Since the system needs to deal with a huge database which consists of an increasing users and images, both effectiveness and efficiency are required. In order to deal with this problem, we propose a face annotation classifier which adopts an online learning and social decomposition approach. The proposed method is shown to have comparable accuracy and better efficiency than that of the widely used Support Vector Machine. Consequently, the proposed framework can reduce the user's tedious efforts to annotate face images and provides a fast response to millions of users.

마이크로컨트롤러의 메모리 구조를 고려한 다층 신경망 최적화

최권택(Kwontaeg Choi) 한국정보기술학회 2021 한국정보기술학회논문지 Vol.19 No.11

With the development of the artificial intelligence industry, various researches are being conducted on AIoT, a fusion of AI technology and IoT technology. For this purpose, it is difficult to apply a neural network algorithm using high-dimensional features due to the small size of dynamically allocable memory in microcontrollers mainly used for this purpose. In this paper, we propose a neural network structure that stores learning parameters in FLASH memory and optimizes memory usage by exchanging input/output memory addresses in SRAM with a small memory capacity. In the UCI HAR dataset experiment, TensorFlow Lite was unable to compile due to lack of memory, whereas the proposed method was able to create a binary file of the same size and perform a multilayer neural network using only 55% FLASH and 16% SRAM. Although the execution time was doubled, real-time execution was possible in 43 ㎳ in the UCI HAR dataset.

생성형 AI 기술을 활용한 미술관 가이드 AI 도슨트

변하정(HajeongByeon),최권택(Kwontaeg Choi) 한국정보기술학회 2023 Proceedings of KIIT Conference Vol.2023 No.11

더미 클래스를 가지는 열린 집합 얼굴 인식 방법의 유효성 검증에 대한 연구

안정호(Jung-Ho Ahn),최권택(KwonTaeg Choi) 한국디지털콘텐츠학회 2017 한국디지털콘텐츠학회논문지 Vol.18 No.3

The open set recognition method should be used for the cases that the classes of test data are not known completely in the training phase. So it is required to include two processes of classification and the validation test. This kind of research is very necessary for commercialization of face recognition modules, but few domestic researches results about it have been published. In this paper, we propose an open set face recognition method that includes two sequential validation phases. In the first phase, with dummy classes we perform classification based on sparse representation. Here, when the test data is classified into a dummy class, we conclude that the data is invalid. If the data is classified into one of the regular training classes, for second validation test we extract four features and apply them for the proposed decision function. In experiments, we proposed a simulation method for open set recognition and showed that the proposed validation test outperform SCI of the well-known validation method.

딥러닝과 GPT-3.5를 사용한 비트코인 투자 분석에 대한 고찰

성창규(Changgue Seoung),최권택(Kwontaeg Choi) 한국정보기술학회 2023 Proceedings of KIIT Conference Vol.2023 No.11

ROMP를 이용한 희소 표현 방식 얼굴 인식 방법론

안정호(Jung-Ho Ahn),최권택(KwonTaeg Choi) 한국디지털콘텐츠학회 2017 한국디지털콘텐츠학회논문지 Vol.18 No.2

It is well-known that the face recognition method via sparse representation has been proved very robust and showed good performance. Its weakness is, however, that its time complexity is very high because it should solve L1-minimization problem to find the sparse solution. In this paper, we propose to use the ROMP(Regularized Orthogonal Matching Pursuit) method for the sparse solution, which solves the L2-minimization problem with regularization condition using the greed strategy. In experiments, we shows that the proposed method is comparable to the existing best L1-minimization solver, Homotopy, but is 60 times faster than Homotopy. Also, we proposed C-SCI method for classification. The C-SCI method is very effective since it considers the sparse solution only without reconstructing the test data. It is shown that the C-SCI method is comparable to, but is 5 times faster than the existing best classification method.