• The Journal of Korea Robotics Society
• /
• v.17 no.4
• /
• pp.425-430
• /
• 2022

As robots are considered one of the mainstream digital transformations, robots with machine vision becomes a main area of study providing the ability to check what robots watch and make decisions based on it. However, it is difficult to find a small object in the image mainly due to the flaw of the most of visual recognition networks. Because visual recognition networks are mostly convolution neural network which usually consider local features. So, we make a model considering not only local feature, but also global feature. In this paper, we propose a detection method of a small marker on the object using deep learning and an algorithm that considers global features by combining Transformer's self-attention technique with a convolutional neural network. We suggest a self-attention model with new definition of Query, Key and Value for model to learn global feature and simplified equation by getting rid of position vector and classification token which cause the model to be heavy and slow. Finally, we show that our model achieves higher mAP than state of the art model YOLOr.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.22-25
• /
• 2020

본 논문에서는 큰 크기의 심층 신경망을 압축하기위해 네트워크 수준의 가중치 공유방법인 Global Weight 패러다임을 최초로 제시한다. 기존의 가중치 공유방법은 계층별로 가중치를 공유하는 것이 대부분이었다. Global Weight 는 기존 방법과 달리 전체 네트워크에서 가중치를 공유하는 효율적인 방법이다. 우리는 Global Weight 를 사용하여 학습되는 새로운 컨볼루션 연산인 Global Weight Convolution(GWConv)연산과 GWConv를 적용한 Global Weight Networks(GWNet)을 제안한다. CIFAR10 데이터셋에서 실험한 결과 2.18 배 압축에서 85.64%, 3.41 배 압축에서 85.46%의 정확도를 보였다. Global Weight 패러다임은 가중치 공유가 궁극적으로 풀고자 했던 중복되는 가중치를 최소화하는 획기적인 방법이며, 추후 심도 있는 연구가 수행될 수 있음을 시사한다.

• The Journal of the Korea Contents Association
• /
• v.6 no.3
• /
• pp.28-37
• /
• 2006

In the last decades, inter-networking technologies advanced more rapidly than any other field. Today, the Internet is one of the most important infrastructure to society as becoming an indispensible tool of people and companies. During mid-1990's, developed countries recognized the advanced network as a basic infrastructure for the future science and technology development. They developed national research and education networks for the development of future science and network technology. In this paper, we made a comprehensive review of global research and education network developments. We also made analysis of Korea's activities on advanced network comparing with those of developed nations, then suggested approaches to improve Korean advanced networks.

• Journal of Korea Multimedia Society
• /
• v.25 no.11
• /
• pp.1643-1652
• /
• 2022

Most of the existing machine reading research has used Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) algorithms as networks. Among them, RNN was slow in training, and Question Answering Network (QANet) was announced to improve training speed. QANet is a model composed of CNN and self-attention. CNN extracts semantic and syntactic information well from the local corpus, but there is a limit to extracting the corresponding information from the global corpus. Graph Convolutional Networks (GCN) extracts semantic and syntactic information relatively well from the global corpus. In this paper, to take advantage of this strength of GCN, we propose I-QANet, which changed the CNN of QANet to GCN. The proposed model performed 1.2 times faster than the baseline in the Stanford Question Answering Dataset (SQuAD) dataset and showed 0.2% higher performance in Exact Match (EM) and 0.7% higher in F1. Furthermore, in the Korean Question Answering Dataset (KorQuAD) dataset consisting only of Korean, the learning time was 1.1 times faster than the baseline, and the EM and F1 performance were also 0.9% and 0.7% higher, respectively.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.6 no.3
• /
• pp.179-183
• /
• 2006

ECG consists of various waveforms of electric signals of heat. Datamining can be used for analyzing and classifying the waveforms. Conventional studies classifying electrocardiogram have problems like extraction of distorted characteristics, overfitting, etc. This study classifies electrocardiograms by using BP algorithm and SVM to solve the problems. As results, this study finds that SVM provides an effective prohibition of overfitting in neural networks and guarantees a sole global solution, showing excellence in generalization performance.

• Journal of applied mathematics & informatics
• /
• v.30 no.5_6
• /
• pp.1051-1065
• /
• 2012

In this paper, the global stability and almost periodicity are investigated for generalized Hopfield neural networks with time-varying neutral delays. Some sufficient conditions are obtained for the existence and globally exponential stability of almost periodic solution by employing fixed point theorem and differential inequality techniques. The results of this paper are new and complement previously known results. Finally, an example is given to demonstrate the effectiveness of our results.

• Journal of Computing Science and Engineering
• /
• v.12 no.1
• /
• pp.24-35
• /
• 2018

Deep convolutional neural network (DCNN) is an advanced technology in image recognition. Because of extreme computing resource requirements, DCNN implementation with software alone cannot achieve real-time requirement. Therefore, the need to implement DCNN accelerator hardware is increasing. In this paper, we present a field programmable gate array (FPGA)-based hardware accelerator design of DCNN targeting handwritten Hangul character recognition application. Also, we present design optimization techniques in SDAccel environments for searching the optimal FPGA design space. The techniques we used include memory access optimization and computing unit parallelism, and data conversion. We achieved about 11.19 ms recognition time per character with Xilinx FPGA accelerator. Our design optimization was performed with Xilinx HLS and SDAccel environment targeting Kintex XCKU115 FPGA from Xilinx. Our design outperforms CPU in terms of energy efficiency (the number of samples per unit energy) by 5.88 times, and GPGPU in terms of energy efficiency by 5 times. We expect the research results will be an alternative to GPGPU solution for real-time applications, especially in data centers or server farms where energy consumption is a critical problem.

• International Journal of Knowledge Content Development & Technology
• /
• v.10 no.4
• /
• pp.29-46
• /
• 2020

The paper describes the quantitative and qualitative dimensions of artificial neural networks (ANN) in India in the global context. The study is based on research publications data (8260) as covered in the Scopus database during 1999-2018. ANN research in India registered 24.52% growth, averaged 11.95 citations per paper, and contributed 9.77% share to the global ANN research. ANN research is skewed as the top 10 countries account for 75.15% of global output. India ranks as the third most productive country in the world. The distribution of research by type of ANN networks reveals that Feed Forward Neural Network type accounted for the highest share (10.18% share), followed by Adaptive Weight Neural Network (5.38% share), Feed Backward Neural Network (2.54% share), etc. ANN research applications across subjects were the largest in medical science and environmental science (11.82% and 10.84% share respectively), followed by materials science, energy, chemical engineering and water resources (from 6.36% to 9.12%), etc. The Indian Institute of Technology, Kharagpur and the Indian Institute of Technology, Roorkee lead the country as the most productive organizations (with 289 and 264 papers). Besides, the Indian Institute of Technology, Kanpur (33.04 and 2.76) and Indian Institute of Technology, Madras (24.26 and 2.03) lead the country as the most impactful organizations in terms of citation per paper and relative citation index. P. Samui and T.N. Singh have been the most productive authors and G.P.S.Raghava (86.21 and 7.21) and K.P. Sudheer (84.88 and 7.1) have been the most impactful authors. Neurocomputing, International Journal of Applied Engineering Research and Applied Soft Computing topped the list of most productive journals.

• Journal of the Korean Academic Society of Industrial Cluster
• /
• v.2 no.1
• /
• pp.14-27
• /
• 2008

Ousters have become a key focus of interest and analysis over the last decade or so, informed by the work of the Harvard business economist Michael Porter. Recent research, however, suggests that the classic Porterian conception of clusters needs to be rethought. In particular, the idea that clusters are geogaphically bounded and integrated units whose primary link to the outside world is through the export of goods and services to global markets is highly Questionable, if not untenable. Relational approaches to clusters and regional development stress the importance of the wider networks and 'pipelines' through which knowledge is exchanged with key partners and collaborators located outside of the particular cluster in question. Rather than the main external links being those between leading firms and global markets, firms may engage in a range of global relations with collaborators and suppliers. This paper address the challenge of rethinking clusters in the light of the recent emphasis on global networks md connections, drawing on experience from m old industrial region in Western Europe Scotland. In assessing cluster experiences and initiatives in Scotland, I examine the development of the oil and gas and electronics clusters. In conclusion, I suggest that cluster initiatives me only likely to generate lasting benefits for the region in question if there is significant local ownership md control of key industries and clusters.

• The Bulletin of The Korean Astronomical Society
• /
• v.27 no.1
• /
• pp.61-61
• /
• 2002