• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.133-135
• /
• 2020

최근 CNN(Convolutional Neural Network)은 영상 분류, 객체 인식 등 다양한 비전 분야에서 우수한 성능을 보여주고 있으나, CNN 모델의 계산량 및 메모리가 매우 커짐에 따라 모바일 또는 IoT(lnternet of Things) 장치와 같은 저전력 환경에 적용되기에는 제한이 따른다. 따라서, CNN 모델의 임무 성능을 유지하연서 네트워크 모델을 압축하는 기법들이 연구되고 있다. 본 논문에서는 행렬 분해 기술인 저계수행렬 근사(Low-rank approximation)와 CP(Canonical Polyadic) 분해 기법을 결합하여 CNN 모델을 압축하는 기법을 제안한다. 제안하는 기법은 계층의 유형에 상관없이 하나의 행렬분해 기법만을 적용하는 기존의 기법과 달리 압축 성능을 높이기 위하여 CNN의 계층 타입에 따라 두 가지 분해 기법을 선택적으로 적용한다. 제안기법의 성능검증을 위하여 영상 분류 CNN 모델인 VGG-16, ResNet50, 그리고 MobileNetV2 모델 압축에 적용하였고, 모델의 계층 유형에 따라 두 가지의 분해 기법을 선택적으로 적용함으로써 저계수행렬 근사 기법만 적용한 경우 보다 1.5~12.1 배의 동일한 압축율에서 분류 성능이 향상됨을 확인하였다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.6
• /
• pp.1207-1214
• /
• 2019

Technologies of the 4th industrial revolution have been unknowingly seeping into our lives. Many IoT based home security systems are using the convolutional neural network(CNN) as good biometrics to recognize a face and protect home and family from intruders since CNN has demonstrated its excellent ability in image recognition. In this paper, three layouts of CNN for 2D and 2.5D image of small dataset with various input image size and filter size are explored. The simulation results show that the layout of CNN with 50*50 input size of 2.5D image, 2 convolution and max pooling layer, and 3*3 filter size for small dataset of 2.5D image is optimal for a home security system with recognition accuracy of 0.966. In addition, the longest CPU time consumption for one input image is 0.057S. The proposed layout of CNN for a face recognition is suitable to control the actuators in the home security system because a home security system requires good face recognition and short recognition time.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.6
• /
• pp.1153-1160
• /
• 2019

In this paper, we experimented to find out how the number of convolution layers, the size, and the number of kernels affect the CNN. In addition, the general CNN was also tested for analysis and compared with the CNN used in the experiment. The neural networks used for the analysis are based on CNN, and each experimental model is experimented with the number of layers, the size, and the number of kernels at a constant value. All experiments were conducted using two layers of fully connected layers as a fixed. All other variables were tested with the same value. As the result of the analysis, when the number of layers is small, the data variance value is small regardless of the size and number of kernels, showing a solid accuracy. As the number of layers increases, the accuracy increases, but from above a certain number, the accuracy decreases, and the variance value also increases, resulting in a large accuracy deviation. The number of kernels had a greater effect on learning speed than other variables.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.5
• /
• pp.865-870
• /
• 2006

The cellular neural networks have shown a vast computing power for the image processing in spite of the simplicity of its structure. However, it is impossible to implement the CNN hardware which would require the same enormous amount of cells as that of the pixels involved in the practical large image. In this parer, the $5\times5$ CNN hardware and the pre post processor which can be used for processing the real large image with a time-multiplexing scheme are implemented. The implemented $5\times5$ CNN hardware and pre post processor is applied to the edge detection of $256\times256$ lena image to evaluate the performance. The total number of block. By the time-multiplexing process is about 4,000 blocks and to control pulses are needed to perform the pipelined operation or the each block. By the experimental resorts, the implemented $5\times5$ CNN hardware and pre post processor can be used to the real large image processing.

• Journal of Broadcast Engineering
• /
• v.24 no.2
• /
• pp.315-328
• /
• 2019

Image steganalysis is an algorithm that classifies input images into stego images with steganography methods and cover images without steganography methods. Previously, handcrafted feature-based steganalysis methods have been mainly studied. However, CNN-based objects recognition has achieved great successes and CNN-based steganalysis is actively studied recently. Unlike object recognition, CNN-based steganalysis requires preprocessing filters to discriminate the subtle difference between cover images from stego images. Therefore, CNN-based steganalysis studies have focused on developing effective preprocessing filters as well as network structures. In this paper, we compare previous studies in same experimental conditions, and based on the results, we analy ze the performance variation caused by the differences in preprocessing filter and network structure.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.8
• /
• pp.311-318
• /
• 2021

Research has been ongoing to detect ships from offshore photographs for a variety of reasons, including maritime security, identifying international trends, and social scientific research. Due to the development of artificial intelligence, R-CNN models for object detection in photographs and images have emerged, and the performance of object detection has risen dramatically. Ship detection in offshore photographs using the R-CNN model has also begun to apply to satellite photography. However, satellite images project large areas, so various objects such as vehicles, landforms, and buildings are sometimes recognized as ships. In this paper, we propose a novel methodology to improve the performance of ship detection in satellite photographs using R-CNN series models. We separate land and sea via marker-based watershed algorithm and perform morphology operations to specify RoI one more time, then detect vessels using R-CNN family models on specific RoI to reduce typology. Using this method, we could reduce the misdetection rate by 80% compared to using only the Fast R-CNN.

• Journal of the Institute of Convergence Signal Processing
• /
• v.24 no.1
• /
• pp.51-56
• /
• 2023

This study investigates the effect of sEMG sampling frequency on CNN learning performance at Korean finger number recognition application. Since the bigger sampling frequency of sEMG signals generates bigger size of input data and takes longer CNN's learning time. It makes making real-time system implementation more difficult and more costly. Thus, there might be appropriate sampling frequency when collecting sEMG signals. To this end, this work choose five different sampling frequencies which are 1,024Hz, 512Hz, 256Hz, 128Hz and 64Hz and investigates CNN learning performance with sEMG data taken at each sampling frequency. The comparative study shows that all CNN recognized Korean finger number one to five at the accuracy of 100% and CNN with sEMG signals collected at 256Hz sampling frequency takes the shortest learning time to reach the epoch at which korean finger number gestures are recognized at the accuracy of 100%.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.11 no.3
• /
• pp.163-171
• /
• 2016

Lane detection is a widely researched topic. Although simple road detection is easily achieved by previous methods, lane detection becomes very difficult in several complex cases involving noisy edges. To address this, we use a Convolution neural network (CNN) for image enhancement. CNN is a deep learning method that has been very successfully applied in object detection and recognition. In this paper, we introduce a robust lane detection method based on a CNN combined with random sample consensus (RANSAC) algorithm. Initially, we calculate edges in an image using a hat shaped kernel, then we detect lanes using the CNN combined with the RANSAC. In the training process of the CNN, input data consists of edge images and target data is images that have real white color lanes on an otherwise black background. The CNN structure consists of 8 layers with 3 convolutional layers, 2 subsampling layers and multi-layer perceptron (MLP) of 3 fully-connected layers. Convolutional and subsampling layers are hierarchically arranged to form a deep structure. Our proposed lane detection algorithm successfully eliminates noise lines and was found to perform better than other formal line detection algorithms such as RANSAC

• Journal of Korea Society of Industrial Information Systems
• /
• v.23 no.4
• /
• pp.41-53
• /
• 2018

Recently, the development of machine learning technology has led to the application of deep learning technology to existing image based application systems. In this context, some researches have been made to apply CNN (Convolutional Neural Network) to the field of fire detection. To verify the effects of existing preprocessing and feature extraction methods on fire detection when combined with CNN, in this paper, the recognition performance and learning time are evaluated by changing the VGG19 CNN structure while gradually increasing the convolution layer. In general, the accuracy is better when the image is not preprocessed. Also it's shown that the preprocessing method and the feature extraction method have many benefits in terms of learning speed.

• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• pp.818-833
• /
• 2021

To address the problems of low precision rate, insufficient feature extraction, and poor contextual ability in existing text sentiment analysis methods, a mixed model account of a CNN-BiLSTM-TE (convolutional neural network, bidirectional long short-term memory, and topic extraction) model was proposed. First, Chinese text data was converted into vectors through the method of transfer learning by Word2Vec. Second, local features were extracted by the CNN model. Then, contextual information was extracted by the BiLSTM neural network and the emotional tendency was obtained using softmax. Finally, topics were extracted by the term frequency-inverse document frequency and K-means. Compared with the CNN, BiLSTM, and gate recurrent unit (GRU) models, the CNN-BiLSTM-TE model's F1-score was higher than other models by 0.0147, 0.006, and 0.0052, respectively. Then compared with CNN-LSTM, LSTM-CNN, and BiLSTM-CNN models, the F1-score was higher by 0.0071, 0.0038, and 0.0049, respectively. Experimental results showed that the CNN-BiLSTM-TE model can effectively improve various indicators in application. Lastly, performed scalability verification through a takeaway dataset, which has great value in practical applications.