• 센서학회지
• /
• 제30권5호
• /
• pp.290-294
• /
• 2021

Experts have designed popular and successful model architectures, which, however, were not the optimal option for different scenarios. Despite the remarkable performances achieved by deep neural networks, manually designed networks for classification tasks are the backbone of object detection. One major challenge is the ImageNet pre-training of the search space representation; moreover, the searched network incurs huge computational cost. Therefore, to overcome the obstacle of the pre-training process, we introduce a network adaptation technique using a pre-trained backbone model tested on ImageNet. The adaptation method can efficiently adapt the manually designed network on ImageNet to the new object-detection task. Neural architecture search (NAS) is adopted to adapt the architecture of the network. The adaptation is conducted on the MobileNetV2 network. The proposed NAS is tested using SSDLite detector. The results demonstrate increased performance compared to existing network architecture in terms of search cost, total number of adder arithmetics (Madds), and mean Average Precision(mAP). The total computational cost of the proposed NAS is much less than that of the State Of The Art (SOTA) NAS method.

• The Journal of the Acoustical Society of Korea
• /
• 제22권2E호
• /
• pp.49-55
• /
• 2003

The modified neural net pattern recognition equations were attempted to apply to speech recognition. The proposed method has a dynamic process of self-organization that has been proved to be successful in recognizing a depth perception in stereoscopic vision. This study has shown that the process has also been useful in recognizing human speech. In the processing, input vocal signals are first compared with standard models to measure similarities that are then given to a process of self-organization in neural net equations. The competitive and cooperative processes are conducted among neighboring input similarities, so that only one winner neuron is finally detected. In a comparative study, it showed that the proposed neural networks outperformed the conventional HMM speech recognizer under the same conditions.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1991년도 한국자동제어학술회의논문집(국내학술편); KOEX, Seoul; 22-24 Oct. 1991
• /
• pp.1028-1033
• /
• 1991

This paper describes the design of a neural network estimator to estimate weld pool sizes for on-line use of quality monitoring and control in GMA welding processes. The estimator utilizes surface temperatures measured at various points on the top surface of the weldment as its input. The main task of the neural net is to realize the mapping characteristics from the point temperatures to the weld pool sizes through training, A series of bead-on plate welding experiments were performed to assess the performance of the neural estimator.

• IEIE Transactions on Smart Processing and Computing
• /
• 제4권1호
• /
• pp.35-43
• /
• 2015

Over the past couple of years, tremendous progress has been made in applying deep learning (DL) techniques to computer vision. Especially, deep convolutional neural networks (DCNNs) have achieved state-of-the-art performance on standard recognition datasets and tasks such as ImageNet Large-Scale Visual Recognition Challenge (ILSVRC). Among them, GoogLeNet network which is a radically redesigned DCNN based on the Hebbian principle and scale invariance set the new state of the art for classification and detection in the ILSVRC 2014. Since there exist various deep learning techniques, this review paper is focusing on techniques directly related to DCNNs, especially those needed to understand the architecture and techniques employed in GoogLeNet network.

• Nuclear Engineering and Technology
• /
• 제54권12호
• /
• pp.4751-4758
• /
• 2022

Two-phase flow may almost exist in every branch of the energy industry. For the corresponding engineering design, it is very essential and crucial to monitor flow patterns and their transitions accurately. With the high-speed development and success of deep learning based on convolutional neural network (CNN), the study of flow pattern identification recently almost focused on this methodology. Additionally, the photographing technique has attractive implementation features as well, since it is normally considerably less expensive than other techniques. The development of such a two-phase flow pattern online monitoring system is the objective of this work, which seldom studied before. The ongoing preliminary engineering design (including hardware and software) of the system are introduced. The flow pattern identification method based on CNNs and transfer learning was discussed in detail. Several potential CNN candidates such as ALexNet, VggNet16 and ResNets were introduced and compared with each other based on a flow pattern dataset. According to the results, ResNet50 is the most promising CNN network for the system owing to its high precision, fast classification and strong robustness. This work can be a reference for the online monitoring system design in the energy system.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• 제25권2호
• /
• pp.26-38
• /
• 2021

Image coloration refers to adding plausible colors to a grayscale image or video. Image coloration has been used in many modern fields, including restoring old photographs, as well as reducing the time spent painting cartoons. In this paper, a method is proposed for colorizing grayscale images using a convolutional neural network. We propose an encoder-decoder model, adapting FusionNet to our purpose. A proper loss function is defined instead of the MSE loss function to suit the purpose of coloring. The proposed model was verified using the ImageNet dataset. We quantitatively compared several colorization models with ours, using the peak signal-to-noise ratio (PSNR) metric. In addition, to qualitatively evaluate the results, our model was applied to images in the test dataset and compared to images applied to various other models. Finally, we applied our model to a selection of old black and white photographs.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 The Third Asian Fuzzy Systems Symposium
• /
• pp.399-403
• /
• 1998

In this paper, we proposed a new fuzzy supervised learning algorithm. We construct, and train, a new type fuzzy neural net to model the linear activation function. Properties of our fuzzy neural net include : (1) a proposed linear activation function ; and (2) a modified delta rule for learning algorithm. We applied this proposed learning algorithm to exclusive OR,3 bit parity using benchmark in neural network and pattern recognition problems, a kind of image recognition.

• 한국멀티미디어학회논문지
• /
• 제22권6호
• /
• pp.665-675
• /
• 2019

In this paper, we compare and analyze the classification performance of deep learning algorithm Convolutional Neural Network(CNN) ac cording to ensemble generation and combining techniques. We used several CNN models(VGG16, VGG19, DenseNet121, DenseNet169, DenseNet201, ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, GoogLeNet) to create 10 ensemble generation combinations and applied 6 combine techniques(average, weighted average, maximum, minimum, median, product) to the optimal combination. Experimental results, DenseNet169-VGG16-GoogLeNet combination in ensemble generation, and the product rule in ensemble combination showed the best performance. Based on this, it was concluded that ensemble in different models of high benchmarking scores is another way to get good results.

• 한국정보처리학회논문지
• /
• 제6권8호
• /
• pp.2098-2115
• /
• 1999

신경망 회로의 해석에서 아직 해결하지 못하는 부분이 은닉층(hidden layer)의 해석이다. 본 논문에서는 신경망 회로의 기본적인 구성회로로써 하나의 입력(p levels)과 하나의 출력(q levels)을 갖는 2-layer Core-Net를 정의하고, 이 Core-Net의 처리 가능 용량(the capacity)은 2차원 무게값 공간(weight space)을 나눌 수 있는 영역의 수로, {{{{ {a}_{p,q} = {{q}^{2}} over {2}p(p-1)- { q} over {2 } (3 { p}^{2 } -7p+2)+ { p}^{2 }-3p+2}}}}임을 수학적 귀납법으로 증명하였다. 이 Core-Net로 신경망 회로의 중간층 해석이 가능함을 시뮬레이션 예제를 통하여 보였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1991년도 한국자동제어학술회의논문집(국내학술편); KOEX, Seoul; 22-24 Oct. 1991
• /
• pp.996-1001
• /
• 1991

We consider the object recognition of industrial tools which have one degree of freedom. In the case of pliers, the shape varies as the jaw angle varies. Thus, a feature vector made from the boundary image also varies along with the jaw angle. But a pattern recognizer should have the ability of classifying objects without any regards to the angle variation. For a pattern recognizer we have utilized a backpropagation neural net. Feature vectors were made from Fourier descriptors of boundary images by truncating the high frequency components, and they were used as inputs to the neural net for training and recognition. In our experiments, backpropagation neural net outperforms the minimum distance rule which is widely used in the pattern recognition. The performance comparison also made under noisy environments.