• Journal of the Korea Society of Computer and Information
• /
• v.21 no.1
• /
• pp.1-8
• /
• 2016

Convolutional neural network (CNN) consists of a few pairs of both convolution layer and subsampling layer. Thus it has more hidden layers than multi-layer perceptron. With the increased layers, the size of convolution mask ultimately determines the total number of weights in CNN because the mask is shared among input images. It also is an important learning factor which makes or breaks CNN's learning. Therefore, this paper proposes the best method to choose the convolution size and the number of layers for learning CNN successfully. Through our face recognition with vast learning examples, we found that the best size of convolution mask is 5 by 5 and 7 by 7, regardless of the number of layers. In addition, the CNN with two pairs of both convolution and subsampling layer is found to make the best performance as if the multi-layer perceptron having two hidden layers does.

• Journal of Korea Multimedia Society
• /
• v.24 no.11
• /
• pp.1552-1559
• /
• 2021

Convolution neural networks (CNNs) show notable performance in image processing and are used as representative core models. CNNs extract and learn features from large amounts of train dataset. In general, it has a structure in which a convolution layer and a fully connected layer are stacked. The core of CNN is the convolution layer. The size of the kernel used for feature extraction and the number that affect the depth of the feature map determine the amount of weight parameters of the CNN that can be learned. These parameters are the main causes of increasing the computational complexity and memory usage of the entire neural network. The most computationally expensive components in CNNs are fully connected and spatial convolution computations. In this paper, we propose a Fourier Convolution Neural Network that performs the operation of the convolution layer in the Fourier domain. We work on modifying and improving the amount of computation by applying the fast fourier transform method. Using the MNIST dataset, the performance was similar to that of the general CNN in terms of accuracy. In terms of operation speed, 7.2% faster operation speed was achieved. An average of 19% faster speed was achieved in experiments using 1024x1024 images and various sizes of kernels.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.1
• /
• pp.133-138
• /
• 2020

This paper introduces a method for improving accuracy using the first convolution layer, which is not used in most modified CNN(: Convolution Neural Networks). In CNN, such as GoogLeNet and DenseNet, the first convolution layer uses only the traditional methods(3×3 convolutional computation, batch normalization, and activation functions), replacing this with RGB-csb. In addition to the results of preceding studies that can improve accuracy by applying RGB values to feature maps, the accuracy is compared with existing CNN using a limited number of images. The method proposed in this paper shows that the smaller the number of images, the greater the learning accuracy deviation, the more unstable, but the higher the accuracy on average compared to the existing CNN. As the number of images increases, the difference in accuracy between the existing CNN and the proposed method decreases, and the proposed method does not seem to have a significant effect.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.11a
• /
• pp.799-802
• /
• 2017

We attempt to combines concepts of Morphological Operator(MO) and Convolutional Neural Networks(CNN) to improve image-to-image translation. To do this, we propose an operation that approximates morphological operations. Also we propose S-Convolution, an operation that extends the operation to use multiple filters like CNN. The experiment result shows that it can learn MO with big filter using multiple S-convolution layer of small filter. To validate effectiveness of the proposed layer in image-to-image translation we experiment with GAN with S-convolution applied. The result showed that GAN with S-convolution can achieve distinct result from that of GAN with CNN.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2019.07a
• /
• pp.281-284
• /
• 2019

본 논문은 CNN (Convolution Neural Networks)의 첫 번째 컨볼루션층(convolution layer)을 RGB-csb(RGB channel separation block)로 대체하여 입력 영상의 RGB 값을 특징 맵에 적용시켜 정확성을 제고시킬 수 있는 선행연구 결과에 추가적으로, 훈련 및 시험 영상 수에 따른 분석을 통하여 정확도 향상방법을 제안한다. 제안한 방법은 영상의 개수가 작을수록 각 학습 간의 정확도 편차가 크게 나타나는 불안정성은 있지만 기존 CNN모델에 비하여 정확도 차이가 증가함을 알 수 있다.

• The Journal of the Acoustical Society of Korea
• /
• v.36 no.6
• /
• pp.407-412
• /
• 2017

In this paper, we propose a speech emotion recognition method using a deep neural network based on the attention mechanism. The proposed method consists of a combination of CNN (Convolution Neural Networks), GRU (Gated Recurrent Unit), DNN (Deep Neural Networks) and attention mechanism. The spectrogram of the speech signal contains characteristic patterns according to the emotion. Therefore, we modeled characteristic patterns according to the emotion by applying the tuned Gabor filters as convolutional filter of typical CNN. In addition, we applied the attention mechanism with CNN and FC (Fully-Connected) layer to obtain the attention weight by considering context information of extracted features and used it for emotion recognition. To verify the proposed method, we conducted emotion recognition experiments on six emotions. The experimental results show that the proposed method achieves higher performance in speech emotion recognition than the conventional methods.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.2
• /
• pp.118-128
• /
• 2018

In this paper, we have studied how to search an underwater object by learning the image generated by the side scan sonar in the convolution neural network. In the method of human side analysis of the side scan image or the image, the convolution neural network algorithm can enhance the efficiency of the analysis. The image data of the side scan sonar used in the experiment is the public data of NSWC (Naval Surface Warfare Center) and consists of four kinds of synthetic underwater objects. The convolutional neural network algorithm is based on Faster R-CNN (Region based Convolutional Neural Networks) learning based on region of interest and the details of the neural network are self-organized to fit the data we have. The results of the study were compared with a precision-recall curve, and we investigated the applicability of underwater object detection in convolution neural networks by examining the effect of change of region of interest assigned to sonar image data on detection performance.

• Korean Journal of Remote Sensing
• /
• v.36 no.5_2
• /
• pp.961-973
• /
• 2020

In this manuscript, a new pansharpening model based on Convolutional Neural Network (CNN) was developed. Dilated convolution, which is one of the representative convolution technologies in CNN, was applied to the model by making it deep and complex to improve the performance of the deep learning architecture. Based on the dilated convolution, the residual network is used to enhance the efficiency of training process. In addition, we consider the spatial correlation coefficient in the loss function with traditional L1 norm. We experimented with Dilated Residual Networks (DRNet), which is applied to the structure using only a panchromatic (PAN) image and using both a PAN and multispectral (MS) image. In the experiments using KOMPSAT-3A, DRNet using both a PAN and MS image tended to overfit the spectral characteristics, and DRNet using only a PAN image showed a spatial resolution improvement over existing CNN-based models.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.10
• /
• pp.935-943
• /
• 2017

CNN (Convolution neural network), which is used for image classification and speech recognition among neural networks learning based on positive data, has been continuously developed to have a high performance structure to date. There are many difficulties to utilize in an embedded system with limited resources. Therefore, we use GPU (General-Purpose Computing on Graphics Processing Units), which is used for general-purpose operation of GPU to solve the problem because we use pre-learned weights but there are still limitations. Since CNN performs simple and iterative operations, the computation speed varies greatly depending on the thread allocation and utilization method in the Single Instruction Multiple Thread (SIMT) based GPGPU. To solve this problem, there is a thread that needs to be relaxed when performing Convolution and Pooling operations with threads. The remaining threads have increased the operation speed by using the method used in the following feature maps and kernel calculations.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.5
• /
• pp.177-183
• /
• 2022

Recently, interesting of home training is getting bigger due to COVID-19. Accordingly, research on applying HAR(human activity recognition) technology to home training has been conducted. However, existing paper of HAR proposed static activity instead of dynamic activity. In this paper, the deep learning model where dynamic exercise posture can be analyzed and the accuracy of the user's exercise posture can be shown is proposed. Fitness images of AI-hub are analyzed by blaze pose. The experiment is compared with three types of deep learning model: RNN(recurrent neural network), LSTM(long short-term memory), CNN(convolution neural network). In simulation results, it was shown that the f1-score of RNN, LSTM and CNN is 0.49, 0.87 and 0.98, respectively. It was confirmed that CNN is more suitable for human activity recognition than other models from simulation results. More exercise postures can be analyzed using a variety learning data.