• Journal of Korean Tunnelling and Underground Space Association
• /
• v.21 no.3
• /
• pp.419-432
• /
• 2019

Most of deep learning model training was proceeded by supervised learning, which is to train labeling data composed by inputs and corresponding outputs. Labeling data was directly generated manually, so labeling accuracy of data is relatively high. However, it requires heavy efforts in securing data because of cost and time. Additionally, the main goal of supervised learning is to improve detection performance for 'True Positive' data but not to reduce occurrence of 'False Positive' data. In this paper, the occurrence of unpredictable 'False Positive' appears by trained modes with labeling data and 'True Positive' data in monitoring of deep learning-based CCTV accident detection system, which is under operation at a tunnel monitoring center. Those types of 'False Positive' to 'fire' or 'person' objects were frequently taking place for lights of working vehicle, reflecting sunlight at tunnel entrance, long black feature which occurs to the part of lane or car, etc. To solve this problem, a deep learning model was developed by simultaneously training the 'False Positive' data generated in the field and the labeling data. As a result, in comparison with the model that was trained only by the existing labeling data, the re-inference performance with respect to the labeling data was improved. In addition, re-inference of the 'False Positive' data shows that the number of 'False Positive' for the persons were more reduced in case of training model including many 'False Positive' data. By training of the 'False Positive' data, the capability of field application of the deep learning model was improved automatically.

• Korean Journal of Acupuncture
• /
• v.33 no.3
• /
• pp.102-113
• /
• 2016

Objectives : This study is aimed at developing and discussing the prediction model of blood stasis pattern of traditional Korean medicine(TKM) using machine learning algorithms: multiple logistic regression and decision tree model. Methods : First, we reviewed the blood stasis(BS) questionnaires of Korean, Chinese, and Japanese version to make a integrated BS questionnaire of patient-reported outcomes. Through a human subject research, patients-reported BS symptoms data were acquired. Next, experts decisions of 5 Korean medicine doctor were also acquired, and supervised learning models were developed using multiple logistic regression and decision tree. Results : Integrated BS questionnaire with 24 items was developed. Multiple logistic regression models with accuracy of 0.92(male) and 0.95(female) validated by 10-folds cross-validation were constructed. By decision tree modeling methods, male model with 8 decision node and female model with 6 decision node were made. In the both models, symptoms of 'recent physical trauma', 'chest pain', 'numbness', and 'menstrual disorder(female only)' were considered as important factors. Conclusions : Because machine learning, especially supervised learning, can reveal and suggest important or essential factors among the very various symptoms making up a pattern identification, it can be a very useful tool in researching diagnostics of TKM. With a proper patient-reported outcomes or well-structured database, it can also be applied to a pre-screening solutions of healthcare system in Mibyoung stage.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.7
• /
• pp.70-81
• /
• 1997

In this paper, we propose HMCNN(hybrid multiple component neural networks) that enhance performance of MCNN by adapting new pattern partitioning algorithm which can cluster many input patterns efficiently. Added neural network performs similar learning procedure that of kohonen network. But it dynamically determine it's number of output neurons using algorithms that decide self-organized number of clusters and patterns in a cluster. The proposed network can effectively be applied to problems of large data as well as huge networks size. As a sresutl, proposed pattern partitioning network can enhance performance results and solve weakness of MCNN like generalization capability. In addition, we can get more fast speed by performing parallel learning than that of other supervised learning networks.

• The Journal of Korea Robotics Society
• /
• v.18 no.3
• /
• pp.346-351
• /
• 2023

With the release of numerous open driving datasets, the demand for domain adaptation in perception tasks has increased, particularly when transferring knowledge from rich datasets to novel domains. However, it is difficult to solve the change 1) in the sensor domain caused by heterogeneous LiDAR sensors and 2) in the environmental domain caused by different environmental factors. We overcome domain differences in the semi-supervised setting with 3-stage model parameter training. First, we pre-train the model with the source dataset with object scaling based on statistics of the object size. Then we fine-tine the partially frozen model weights with copy-and-paste augmentation. The 3D points in the box labels are copied from one scene and pasted to the other scenes. Finally, we use the knowledge distillation method to update the student network with a moving average from the teacher network along with a self-training method with pseudo labels. Test-Time Augmentation with varying z values is employed to predict the final results. Our method achieved 3rd place in ECCV 2022 workshop on the 3D Perception for Autonomous Driving challenge.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2001.11a
• /
• pp.226-230
• /
• 2001

오류 역전파 알고리즘을 영상 인식에 적용한 경우 은닉층의 노드 수를 경험적으로 설정하여야 하는 문제점이 있다. 따라서 본 논문에서는 오류 역전파 알고리즘의 은닉층 노드 수를 동적으로 설정하는 문제를 해결하기 위해 ART1을 수정하여 지도 학습 방법과 결합한 자가 생성 지도 학습 알고리즘을 제안하였다. 제안된 학습 알고리즘의 성능을 평가하기 위해 콘테이너 영상의 문자 및 숫자 인식 문제에 적용하여 기존의 오류 역전파 알고리즘과 성능을 비교, 분석하였다. 실험 결과에서는 제안된 자가 생성 지도 학습알고리즘이 기존의 오류 역전과 알고리즘보다 지역 최소화에 빠질 가능성이 감소하였으며 학습 시간과 수렴성이 개선되었을 뿐만 아니라, 영상 인식에 적용할 수 있는 가능성도 제시하였다.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.5
• /
• pp.85-91
• /
• 1989

In this paper, we propose a neural net system for speech recognition, which is composed of two neural networks. Firstly the self-supervised BP(Back Propagation) network generates the distributed concept corresponding to the activity pattern in the hidden units. And then the self-organizing neural network forms a concept map which directly displays the similarity relations between concepts. By doing the above, the difficulty in learning the conventional BP network is solved and the weak side of BP falling into a pattern matcher is gone, while the strong point of generating the various internal representations is used. And we have obtained the concept map which is more orderly than the Kohonen's SOFM. The proposed neural net system needs not any special preprocessing and has a self-learning ability.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.18 no.8
• /
• pp.1939-1951
• /
• 1994

This study introduces a new tool breakage detecting technology comprised of an unsupervised neural network combined with adaptive time series autoregressive(AR) model where parameters are estimated recursively at each sampling instant using a parameter adaptation algorithm based on an RLS(Recursive Least Square). Experiment indicates that AR parameters are good features for tool breakage, therefore it can be detected by tracking the evolution of the AR parameters during milling process. an ART 2(Adaptive Resonance Theory 2) neural network is used for clustering of tool states using these parameters and the network is capable of self organizing without supervised learning. This system operates successfully under the wide range of cutting conditions without a priori knowledge of the process, with fast monitoring time.

• Journal of Korea Multimedia Society
• /
• v.23 no.2
• /
• pp.135-145
• /
• 2020

Recently, deep-learning based methods for low-light image enhancement accomplish great success through supervised learning. However, they still suffer from the lack of sufficient training data due to difficulty of obtaining a large amount of low-/normal-light image pairs in real environments. In this paper, we propose an unsupervised learning approach for single low-light image enhancement using the bright channel prior (BCP), which gives the constraint that the brightest pixel in a small patch is likely to be close to 1. With this prior, pseudo ground-truth is first generated to establish an unsupervised loss function. The proposed enhancement network is then trained using the proposed unsupervised loss function. To the best of our knowledge, this is the first attempt that performs a low-light image enhancement through unsupervised learning. In addition, we introduce a self-attention map for preserving image details and naturalness in the enhanced result. We validate the proposed method on various public datasets, demonstrating that our method achieves competitive performance over state-of-the-arts.

• Smart Structures and Systems
• /
• v.30 no.6
• /
• pp.557-569
• /
• 2022

Vibration-based structural health monitoring (SHM) is crucial for the dynamic maintenance of civil building structures to protect property security and the lives of the public. Analyzing these vibrations with modern artificial intelligence and deep learning (DL) methods is a new trend. This paper proposed an unsupervised deep learning method based on a convolutional autoencoder (CAE), which can overcome the limitations of conventional supervised deep learning. With the convolutional core applied to the DL network, the method can extract features self-adaptively and efficiently. The effectiveness of the method in detecting damage is then tested using a benchmark model. Thereafter, this method is used to detect damage and instant disaster events in a rubber bearing-isolated gymnasium structure. The results indicate that the method enables the CAE network to learn the intact vibrations, so as to distinguish between different damage states of the benchmark model, and the outcome meets the high-dimensional data distribution characteristics visualized by the t-SNE method. Besides, the CAE-based network trained with daily vibrations of the isolating layer in the gymnasium can precisely recover newly collected vibration and detect the occurrence of the ground motion. The proposed method is effective at identifying nonlinear variations in the dynamic responses and has the potential to be used for structural condition assessment and safety warning.

• Annual Conference on Human and Language Technology
• /
• 1995.10a
• /
• pp.39-42
• /
• 1995

문장 안에서 한 단어가 가지는 올바른 의미를 얻기 위해 모듈화된 신경망을 이용하였다. 앞부분에 놓인 신경망은 코호넨 신경망으로 사용자의 지도가 개입되지 않은 상태로 자율학습(Unsupervised learning)이 이루어지고, 뒤에 놓인 신경망은 앞에서 결과로 얻은 2차원의 자기 조직화 형상지도(Self-organizing feature map)를 바탕으로 역전파 신경망을 이용한 지도학습(Supervised learning)을 하게 하였다. 입력 자료는 구문분석된 문장의 조사 정보를 활용하여 입력 위치를 정해준 명사의 의미표지와 동사의 의미표지를 사용하였다. 중의성이 있는 단어를 가지는 문장은 중의성의 가지수 만큼 테스트 입력 자료가 되어 신경망을 통과하여 의미를 결정하도록 한다.