• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.12
• /
• pp.1581-1587
• /
• 2020

With the emergence of convolutional neural network in the field of machine learning, the model for solving image processing problems has seen rapid development. However, the computing resources required are also rising, making it difficult to learn from a typical environment. Attention mechanism is originally proposed to prevent the gradient vanishing problem of the recurrent neural network, but this can also be used in a direction favorable to learning of the convolutional neural network. In this paper, attention mechanism is applied to convolutional neural network, and the excellence of the proposed method is demonstrated through the comparison of learning time and performance difference at this time. The proposed model showed that both learning time and performance were superior in object detection based on YOLO compared to models without attention mechanism, and experimentally demonstrated that learning time could be significantly reduced. In addition, this is expected to increase accessibility to machine learning by end users.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.05a
• /
• pp.398-401
• /
• 2018

본 연구에서는 심충 합성곱 신경망(Deep CNN)과 Connectionist Temporal Classification (CTC) 알고리즘을 사용하여 강제정렬 (force-alignment)이 이루어진 코퍼스 없이도 학습이 가능한 음소 인식 모델을 제안한다. 최근 해외에서는 순환 신경망(RNN)과 CTC 알고리즘을 사용한 딥 러닝 기반의 음소 인식 모델이 활발히 연구되고 있다. 하지만 한국어 음소 인식에는 HMM-GMM 이나 인공 신경망과 HMM 을 결합한 하이브리드 시스템이 주로 사용되어 왔으며, 이 방법 은 최근의 해외 연구 사례들보다 성능 개선의 여지가 적고 전문가가 제작한 강제정렬 코퍼스 없이는 학습이 불가능하다는 단점이 있다. 또한 RNN 은 학습 데이터가 많이 필요하고 학습이 까다롭다는 단점이 있어, 코퍼스가 부족하고 기반 연구가 활발하게 이루어지지 않은 한국어의 경우 사용에 제약이 있다. 이에 본 연구에서는 강제정렬 코퍼스를 필요로 하지 않는 CTC 알고리즘을 도입함과 동시에, RNN 에 비해 더 학습 속도가 빠르고 더 적은 데이터로도 학습이 가능한 합성곱 신경망(CNN)을 사용하여 딥 러닝 모델을 구축하여 한국어 음소 인식을 수행하여 보고자 하였다. 이 모델을 통해 본 연구에서는 한국어에 존재하는 49 가지의 음소를 추출하는 세 종류의 음소 인식기를 제작하였으며, 최종적으로 선정된 음소 인식 모델의 PER(phoneme Error Rate)은 9.44 로 나타났다. 선행 연구 사례와 간접적으로 비교하였을 때, 이 결과는 제안하는 모델이 기존 연구 사례와 대등하거나 조금 더 나은 성능을 보인다고 할 수 있다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.11a
• /
• pp.11-12
• /
• 2017

본 논문에서는 부류가 정해진 훈련 집합이 불필요한 온라인 학습 기반 추적 기법을 제안한다. 추적기의 학습을 위해 합성곱 신경망(convolutional neural network: CNN)을 이용하였다. 추적영상으로부터 직접 훈련 샘플을 수집함으로써 분류기 학습을 위한 비용을 감소시킬 수 있었고, 목표 영상에 적응적인 객체 모델을 생성할 수 있다. 실험 결과를 통해 제안하는 방법이 우수한 성능을 보임을 입증하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.454-460
• /
• 2020

Artificial intelligence assistants that provide speech recognition operate through cloud-based voice recognition with high accuracy. In cloud-based speech recognition, Wake-Up-Word (WUW) detection plays an important role in activating devices on standby. In this paper, we compare the performance of Convolutional Neural Network (CNN)-based WUW detection models for mobile devices by using Google's speech commands dataset, using the spectrogram and mel-frequency cepstral coefficient features as inputs. The CNN models used in this paper are multi-layer perceptron, general convolutional neural network, VGG16, VGG19, ResNet50, ResNet101, ResNet152, MobileNet. We also propose network that reduces the model size to 1/25 while maintaining the performance of MobileNet is also proposed.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.582-585
• /
• 2024

이 연구는 컴퓨팅 자원이 제한된 환경에서 딥러닝 모델의 문제를 해결하기 위해 합성곱 신경망(CNN)에서 동적 가지치기 모델의 적용을 탐구한다. 첫째, 동적 가지치기 모델의 원리와 방법에 대해 기존 방법과의 비교를 소개한다. 둘째, 기존적인 방법 동적 가지치기 모델의 구현 과정 및 결과 분석을 포함한 실험 단계를 자세히 설명한다. 실험 결과는 동적 가지치기 모델이 적절한 훈련에서 모델의 분류 성능을 효과적으로 향상시킬 수 있으며 강력한 일반화 능력을 가지고 있음을 보여준다. 마지막으로 딥러닝 방법과 기존 방법의 차이점과 장단점을 분석하고 요약하여 실제 적용에서 딥러닝 모델 배치에 유용한 탐색과 참고 자료를 제공한다. 이 연구는 딥러닝 분야에서 동적 가지치기 모델의 적용을 추가로 탐색하기 위한 중요한 이론 및 실습 기반을 제공한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.267-269
• /
• 2020

입모양 인식(Lip Reading) 기술은 입술 움직임을 통해 발화를 분석하는 기술이다. 본 논문에서는 일상적으로 사용하는 10개의 상용구에 대해서 발화자의 안면 움직임 분석을 통해 실시간으로 분류하는 연구를 진행하였다. 시간상의 연속된 순서를 가진 영상 데이터의 특징을 고려하여 3차원 합성곱 신경망 (Convolutional Neural Network)을 사용하여 진행하였지만, 실시간 시스템 구현을 위해 연산량 감소가 필요했다. 이를 해결하기 위해 차 영상을 이용한 2차원 합성곱 신경망과 LSTM 순환 신경망 (Long Short-Term Memory) 결합 모델을 설계하였고, 해당 모델을 이용하여 실시간 시스템 구현에 성공하였다.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.36 no.4
• /
• pp.221-232
• /
• 2023

This paper presents a novel convolutional neural network (CNN)-based approach for predicting bolt clamping force in the early bolt loosening state of bolted structures. The approach entails tightening eight bolts with different clamping forces and generating frequency responses, which are then used to create a similarity map. This map quantifies the magnitude and shape similarity between the frequency responses and the initial model in a fully fastened state. Krylov subspace-based model order reduction is employed to efficiently handle the large amount of frequency response data. The CNN model incorporates a regression output layer to predict the clamping forces of the bolts. Its performance is evaluated by training the network by using various amounts of training data and convolutional layers. The input data for the model are derived from the magnitude and shape similarity map obtained from the frequency responses. The results demonstrate the diagnostic potential and effectiveness of the proposed approach in detecting early bolt loosening. Accurate bolt clamping force predictions in the early loosening state can thus be achieved by utilizing the frequency response data and CNN model. The findings afford valuable insights into the application of CNNs for assessing the integrity of bolted structures.

• Economic and Environmental Geology
• /
• v.53 no.4
• /
• pp.479-489
• /
• 2020

A fault is a geological structure that can be a migration path or a cap rock of hydrocarbon such as oil and gas, formed from source rock. The fault is one of the main targets of seismic exploration to find reservoirs in which hydrocarbon have accumulated. However, conventional fault detection methods using lateral discontinuity in seismic data such as semblance, coherence, variance, gradient magnitude and fault likelihood, have problem that professional interpreters have to invest lots of time and computational costs. Therefore, many researchers are conducting various studies to save computational costs and time for fault interpretation, and machine learning technologies attracted attention recently. Among various machine learning technologies, many researchers are conducting fault interpretation studies using the support vector machine, multi-layer perceptron, deep neural networks and convolutional neural networks algorithms. Especially, researchers use not only their own convolution networks but also proven networks in image processing to predict fault locations and fault information such as strike and dip. In this paper, by investigating and analyzing these studies, we found that the convolutional neural networks based on the U-Net from image processing is the most effective one for fault detection and interpretation. Further studies can expect better results from fault detection and interpretation using the convolutional neural networks along with transfer learning and data augmentation.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.6
• /
• pp.592-599
• /
• 2020

This paper proposes a Convolutional Recurrent Neural Net (CRNN) structure that can simultaneously reflect both static and dynamic characteristics of seismic waveforms for various earthquake events classification. Addressing various earthquake events, including not only micro-earthquakes and artificial-earthquakes but also macro-earthquakes, requires both effective feature extraction and a classifier that can discriminate seismic waveform under noisy environment. First, we extract the static characteristics of seismic waveform through an attention-based convolution layer. Then, the extracted feature-map is sequentially injected as input to a multi-input single-output Long Short-Term Memory (LSTM) network structure to extract the dynamic characteristic for various seismic event classifications. Subsequently, we perform earthquake events classification through two fully connected layers and softmax function. Representative experimental results using domestic and foreign earthquake database show that the proposed model provides an effective structure for various earthquake events classification.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.36 no.6
• /
• pp.469-481
• /
• 2018

DL (Deep Learning) is getting popular in various fields to implement artificial intelligence that resembles human learning and cognition. DL based on complicate structure of the ANN (Artificial Neural Network) requires computing power and computation cost. Variety of DL models with improved performance have been developed with powerful computer specification. The main purpose of this paper is to detect buildings from aerial images and evaluate performance of Mask R-CNN (Region-based Convolutional Neural Network) developed by FAIR (Facebook AI Research) team recently. Mask R-CNN is a R-CNN that is evaluated to be one of the best ANN models in terms of performance for semantic segmentation with pixel-level accuracy. The performance of the DL models is determined by training ability as well as architecture of the ANN. In this paper, we characteristics of the Mask R-CNN with various types of the images and evaluate possibility of the generalization which is the ultimate goal of the DL. As for future study, it is expected that reliability and generalization of DL will be improved by using a variety of spatial information data for training of the DL models.