• Proceedings of the Korea Information Processing Society Conference
• /
• 한국정보처리학회 2023년도 춘계학술발표대회
• /
• pp.510-512
• /
• 2023

In order to enhance a model's capability for detecting facial expressions, this research suggests a pipeline that makes use of the GradCAM component. The patching module and the pseudo-labeling module make up the pipeline. The patching component takes the original face image and divides it into four equal parts. These parts are then each input into a 2Dconvolutional layer to produce a feature vector. Each picture segment is assigned a weight token using GradCAM in the pseudo-labeling module, and this token is then merged with the feature vector using principal component analysis. A convolutional neural network based on transfer learning technique is then utilized to extract the deep features. This technique applied on a public dataset MMI and achieved a validation accuracy of 96.06% which is showing the effectiveness of our method.

• The Journal of Korean Association of Computer Education
• /
• 제12권2호
• /
• pp.13-22
• /
• 2009

In this paper, we propose a new dynamic syllabus composition system to solve the problems of a static syllabus which can appear in the field of computer education, where the relationship between pre-post study subjects is clear and teachers may grasp easily the degree of understanding of learners in real time. Our dynamic syllabus composition system is designed to be adjusted according to the physical change of the amount of education and the level of learners, which is based on the priority of educational objects. The result of instance performed on two groups of different entering behavior shows that the proposed method enhances the degree of transfer of education and helps us teach a class around the representative subject which has to be dealt with in the class, so that it is very effective for the achievement of educational objects prior to others.

• International Journal of Computer Science & Network Security
• /
• 제22권5호
• /
• pp.143-148
• /
• 2022

Brain tumor classification is an important process that allows doctors to plan treatment for patients based on the stages of the tumor. To improve classification performance, various CNN-based architectures are used for brain tumor classification. Existing methods for brain tumor segmentation suffer from overfitting and poor efficiency when dealing with large datasets. The enhanced CNN architecture proposed in this study is based on U-Net for brain tumor segmentation, RefineNet for pattern analysis, and SegNet architecture for brain tumor classification. The brain tumor benchmark dataset was used to evaluate the enhanced CNN model's efficiency. Based on the local and context information of the MRI image, the U-Net provides good segmentation. SegNet selects the most important features for classification while also reducing the trainable parameters. In the classification of brain tumors, the enhanced CNN method outperforms the existing methods. The enhanced CNN model has an accuracy of 96.85 percent, while the existing CNN with transfer learning has an accuracy of 94.82 percent.

• The Journal of the Acoustical Society of Korea
• /
• 제40권5호
• /
• pp.488-495
• /
• 2021

We propose a speech recognition system based on conformer. Conformer is known to be convolution-augmented transformer, which combines transfer model for capturing global information with Convolution Neural Network (CNN) for exploiting local feature effectively. The baseline system is developed to be a transfer-based speech recognition using Long Short-Term Memory (LSTM)-based language model. The proposed system is a system which uses conformer instead of transformer with transformer-based language model. When Electronics and Telecommunications Research Institute (ETRI) speech corpus in AI-Hub is used for our evaluation, the proposed system yields 5.7 % of Character Error Rate (CER) while the baseline system results in 11.8 % of CER. Even though speech corpus is extended into other domain of AI-hub such as NHNdiguest speech corpus, the proposed system makes a robust performance for two domains. Throughout those experiments, we can prove a validation of the proposed system.

• The Journal of Korea Robotics Society
• /
• 제6권4호
• /
• pp.301-307
• /
• 2011

Based on the stability criteria of ZMP (Zero Moment Point), this paper proposes an adjusting algorithm that modifies walking trajectory of a bipedal robot for stable walking by analyzing ZMP trajectory of it. In order to maintain walking balance of the bipedal robot, ZMP should be located within a supporting polygon that is determined by the foot supporting area with stability margin. Initially tilting imposed to the trajectory of the upper body is proposed to transfer ZMP of the given walking trajectory into the stable region for the minimum stability. A neural network method is also proposed for the stable walking trajectory of the biped robot. It uses backpropagation learning with angles and angular velocities of all joints with tilting to get the improved walking trajectory. By applying the optimized walking trajectory that is obtained with the neural network model, the ZMP trajectory of the bipedal robot is certainly located within a stable area of the supporting polygon. Experimental results show that the optimally learned trajectory with neural network gives more stability even though the tilting of the pelvic joint has a great role for walking stability.

• Journal of the Institute of Convergence Signal Processing
• /
• 제12권3호
• /
• pp.212-216
• /
• 2011

Fault detection technique for photovoltaic power systems is significant to dramatically reduce economic damage in industrial fields. This paper presents a novel fault detection approach using Fourier neural networks and stochastic decision making strategy for photovoltaic systems. We achieve neural modeling to represent its nonlinear dynamic behaviors through a gradient descent based learning algorithm. Next, a general likelihood ratio test (GLRT) is derived for constructing a decision malling mechanism in stochastic fault detection. A testbed of photovoltaic power systems is established to conduct real-time experiments in which the DC power line communication (DPLC) technique is employed to transfer data sets measured from the photovoltaic panels to PC systems. We demonstrate our proposed fault detection methodology is reliable and practicable over this real-time experiment.

• The Journal of Korean Association of Computer Education
• /
• 제6권2호
• /
• pp.41-51
• /
• 2003

In the distance education systems, there are limited things while performing educational activities because there are to be problems in structural features of the Web. HTTP, a connection-less protocol, performs requests of client, however, does not hold on the status. Thus, by features of the Web, it is difficult to hold on the connection of learners and trace information asked by learners. Moreover, these problems make impossible not only a learner's connection continuity but also on-line interaction among the learners in the distance education. This thesis developed CHTP, an connection-based hypertext transfer protocol, based on HTTP and a new platform of distance education in order to track activities of learners. The developed web extension platform will make it easy to build up system for being helpful in a distance education because this thesis proposes a standardized way in the protocol.

• Annual Conference on Human and Language Technology
• /
• 한국정보과학회언어공학연구회 2022년도 제34회 한글 및 한국어 정보처리 학술대회
• /
• pp.571-575
• /
• 2022

본 논문에서는 통계청에서 매월 작성되는 온라인쇼핑동향조사를 위해, 언어모델 전이학습 기반 분류모델 학습 방법론을 이용하여, 관세청 제공 전자상거래 수입 목록통관 자료를 처리하기 위해서 해외 직접 구매 상품군 분류 모델을 구축한다. 최근에 텍스트 분류 태스크에서 많이 이용되는 BERT 기반의 언어모델을 이용하며 기존의 색인어 정보 분석 과정이나 사례사전 구축 등의 중간 단계 없이 해외 직접 판매 및 구매 상품군을 94%라는 높은 예측 정확도로 분류가 가능해짐을 알 수 있다.

• Knowledge Management Research
• /
• 제22권3호
• /
• pp.91-106
• /
• 2021

With the recent rapid development of deep learning technology, the demand for analyzing huge text documents in the national R&D field from various perspectives is rapidly increasing. In particular, interest in the application of a BERT(Bidirectional Encoder Representations from Transformers) language model that has pre-trained a large corpus is growing. However, the terminology used frequently in highly specialized fields such as national R&D are often not sufficiently learned in basic BERT. This is pointed out as a limitation of understanding documents in specialized fields through BERT. Therefore, this study proposes a method to build an R&D KoBERT language model that transfers national R&D field knowledge to basic BERT using further pre-training. In addition, in order to evaluate the performance of the proposed model, we performed classification analysis on about 116,000 R&D reports in the health care and information and communication fields. Experimental results showed that our proposed model showed higher performance in terms of accuracy compared to the pure KoBERT model.

• Proceedings of the Korea Information Processing Society Conference
• /
• 한국정보처리학회 2024년도 춘계학술발표대회
• /
• pp.768-771
• /
• 2024

인간의 산업 활동으로 인하여 동물들의 생존이 위협받으면서, 동물의 서식 분포를 효과적으로 파악할 수 있는 자동 야생동물 모니터링 기술의 필요성이 점점 더 커지고 있다. 그중에서도 동물 소리 분류 기술은 시각적으로 식별이 어려운 동물에게도 효과적으로 적용할 수 있는 장점으로 인하여 널리 사용되고 있다. 최근 심층학습 기반의 분류 모델들이 좋은 판별 성능을 보여주고 있어 동물 소리 분류에 많이 사용되고 있지만, 희귀종과 같이 개체 수가 적어 데이터가 부족한 경우에는 학습이 제대로 이루어지지 않을 수 있다. 또한, 이러한 모델들은 모델 내부에서 일어나는 추론 과정을 알 수 없어 결과를 완전히 신뢰하고 사용하는 데 제약이 따른다. 이에 본 논문에서는 전이 학습을 통해 데이터 부족 문제를 고려하고, SHAP을 이용하여 분류 모델의 추론 과정을 해석하는 설명가능한 동물 소리 분류 기법을 제안한다. 실험 결과, 제안하는 기법은 지도 학습을 한 경우보다 분류 성능이 향상됨을 확인하였으며, SHAP 분석을 통해 모델의 분류 근거를 이해할 수 있었다.