• Journal of the Semiconductor & Display Technology
• /
• v.22 no.2
• /
• pp.128-132
• /
• 2023

Recently, with the development of deep learning technology, research on pedestrian attribute recognition technology using deep neural networks has been actively conducted. Existing pedestrian attribute recognition techniques can be obtained in such a way as global-based, regional-area-based, visual attention-based, sequential prediction-based, and newly designed loss function-based, depending on how pedestrian attributes are detected. It is known that the performance of these pedestrian attribute recognition technologies varies greatly depending on the type of backbone network that constitutes the deep neural networks model. Therefore, in this paper, several backbone networks are applied to the baseline pedestrian attribute recognition model and the performance changes of the model are analyzed. In this paper, the analysis is conducted using Resnet34, Resnet50, Resnet101, Swin-tiny, and Swinv2-tiny, which are representative backbone networks used in the fields of image classification, object detection, etc. Furthermore, this paper analyzes the change in time complexity when inferencing each backbone network using a CPU and a GPU.

• Journal of Korean Association for Spatial Structures
• /
• v.9 no.4
• /
• pp.73-80
• /
• 2009

It is very important that predict the inelastic seismic behavior exactly for seismic performance evaluation of a building in the performance based seismic design. But, it is difficulty that predict the building behavior of actual and exact in simplified load-deformation relation of structural material and members. In this study, system ductility and story ductility capacity of building structure used to the Backbone hinge Model are estimated and compared considering the characteristics of load-deformation relation of structural material and members. Analyses results, bilinear hinge model has lower system ductility and story ductility demands than those of backbone hinge model.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.6B
• /
• pp.1165-1173
• /
• 1999

In order to support various multimedia communication services, a well balanced backbone network should be designed using recently advanced optical communication technologies. In this paper an optimal backbone network configuration design is presented fur Korean traffic environment. A new traffic model, Population-Distance-Gross Group Products(PDG) traffic model, is devised. In Korean network traffic environment, six regional centers are selected, link capacities between the regional centers are estimated from the PDG traffic model, and the overall network configuration is designed for the all-optical backbone network in Korea. A simulation study is carried out to verify the desired performance of the designed backbone network. Simulation results show that performance of the backbone network is well balanced to support various communication services in Korea in the 2000s.

• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.346-350
• /
• 2023

Trajectory prediction is an essential element for driving autonomous vehicles, and various trajectory prediction models have emerged with the development of deep learning technology. Convolutional neural network (CNN) is the most commonly used neural network architecture for extracting the features of visual images, and the latest models exhibit high performances. This study was conducted to identify an efficient CNN backbone model among the components of deep learning models for trajectory prediction. We changed the existing CNN backbone network of multiple-trajectory prediction models used as feature extractors to various state-of-the-art CNN models. The experiment was conducted using nuScenes, which is a dataset used for the development of autonomous vehicles. The results of each model were compared using frequently used evaluation metrics for trajectory prediction. Analyzing the impact of the backbone can improve the performance of the trajectory prediction task. Investigating the influence of the backbone on multiple deep learning models can be a future challenge.

• The Journal of Korea Robotics Society
• /
• v.3 no.3
• /
• pp.203-211
• /
• 2008

This work proposes structure of spring backbone micro endoscope. For effective surgery in narrow and limited space, many manipulators are developing that different to existed structure. This device can move like elephant nose or snake unlike the existing robots. For this motion, a mechanism that uses spring backbone and wires has been developed. The new type endoscope that has Z axis motion for spring structure, therefore it has 3 degree of freedom, two rotations and one linear motion. And new kinematics for backbone structure is proposed using simple geographic analysis. The Jacobian and stiffness modeling are also derived. Exact actuator sizing is determined using stiffness model. Finally, the proposed kinematics are verified by simulation and experiments.

• Journal of the Korea Society for Simulation
• /
• v.31 no.1
• /
• pp.11-18
• /
• 2022

Recently, research to classify human activity using imagery has been actively conducted for the purpose of crime prevention and facility safety in public places and facilities. In order to improve the performance of human activity classification, most studies have applied deep learning based-transfer learning. However, despite the increase in the number of backbone network models that are the basis of deep learning as well as the diversification of architectures, research on finding a backbone network model suitable for the purpose of operation is insufficient due to the atmosphere of using a certain model. Thus, this study applies the transfer learning into recently developed deep learning backborn network models to build an intelligent system that classifies human activity using imagery. For this, 12 types of active and high-contact human activities based on sports, not basic human behaviors, were determined and 7,200 images were collected. After 20 epochs of transfer learning were equally applied to five backbone network models, we quantitatively analyzed them to find the best backbone network model for human activity classification in terms of learning process and resultant performance. As a result, XceptionNet model demonstrated 0.99 and 0.91 in training and validation accuracy, 0.96 and 0.91 in Top 2 accuracy and average precision, 1,566 sec in train process time and 260.4MB in model memory size. It was confirmed that the performance of XceptionNet was higher than that of other models.

• International Journal of Naval Architecture and Ocean Engineering
• /
• v.6 no.4
• /
• pp.1160-1181
• /
• 2014

Springing and whipping are becoming increasingly important considerations in ship design as container ships increase in size. In this study, the springing and whipping characteristics of a large container ship were investigated through a series of systematic model tests in waves. A multi-segmented hull model with a backbone was adopted for measurement of springing and whipping signals. A conversion method for extracting torsion springing and whipping is described in this paper for the case of an open-section backbone. Higher-order springing, higher-mode torsion responses, and the effects of linear and nonlinear springing in irregular waves are highlighted in the discussion.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.35 no.3
• /
• pp.45-58
• /
• 2010

Due to rapid developments of IT technologies, convergence services like IPTV (Internet protocol television) are shown up. Even though expert and customer had great expectations for this innovative service, commercialization was delayed for years by the legal dispute between industry players. One of the biggest problems was that whether internet backbone providers have to share their internet network backbone with IPTV service providers (which don't have network backbone) or not. As other countries, Korean government set the rules that ISP have to offer indiscriminate access to other IPTV service provider. At the same time, internet backbone providers can charge access charge based on cost by way of compensation. Thus access charge is very critical to the IPTV industry players. The objective of this paper is to provide model that can calculate the reasonable access charge by system dynamics, based upon real data in Korean telecommunication industry.

• Proceedings of the IEEK Conference
• /
• 2003.11c
• /
• pp.101-104
• /
• 2003

As optical network technology advances, the Dense-Wavelength Division Multiplexing(DWDM) networks have been widely accepted as a promising approach to the Next Generation Optical Internet (NGOI) backbone networks. Especially. a fault/attack management scheme in NGOI backbone networks is one of the most important issues because a short service disruption in DWDM networks carrying extremely high data rates causes loss of vast traffic volumes. In this paper, we suggest a fault/attack management model for NGOI backbone networks and propose a fault/attack recovery procedure in IP/GMPLS over DWDM.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.5B
• /
• pp.889-898
• /
• 2000

In order to support various broadband multimedia servies in the future, we designed a well balanced WDM backbone network. In Korean network traffic environment, six regional centers are selected, link capacities between the regional centers are estimated from the PDI traffic model, and the overall network configuration is designed for the all-optical backbone network. Also, we designed a basic configuration to be able to protect minimum communication capability against link failure. A simulation study is carried out to verify the desired performance of the designed WDM backbone network. Simulation results show that performance of the backbone network is well balanced to support various communication services in Korea in the mid 2000s