• International journal of advanced smart convergence
• /
• v.7 no.2
• /
• pp.95-100
• /
• 2018

In this paper, we propose a binarized multi-scale module to accelerate the speed of the pose estimating deep neural network. Recently, deep learning is also used for fine-tuned tasks such as pose estimation. One of the best performing pose estimation methods is based on the usage of two neural networks where one computes the heat maps of the body parts and the other computes the part affinity fields between the body parts. However, the convolution filtering with a large kernel filter takes much time in this model. To accelerate the speed in this model, we propose to change the large kernel filters with binarized multi-scale modules. The large receptive field is captured by the multi-scale structure which also prevents the dropdown of the accuracy in the binarized module. The computation cost and number of parameters becomes small which results in increased speed performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.3
• /
• pp.1477-1491
• /
• 2017

Inspired by the multi-scale nature of hippocampal place cells, a biologically inspired model based on a multi-scale spatial representation for goal-directed navigation is proposed in order to achieve robotic spatial cognition and autonomous navigation. First, a map of the place cells is constructed in different scales, which is used for encoding the spatial environment. Then, the firing rate of the place cells in each layer is calculated by the Gaussian function as the input of the Q-learning process. The robot decides on its next direction for movement through several candidate actions according to the rules of action selection. After several training trials, the robot can accumulate experiential knowledge and thus learn an appropriate navigation policy to find its goal. The results in simulation show that, in contrast to the other two methods(G-Q, S-Q), the multi-scale model presented in this paper is not only in line with the multi-scale nature of place cells, but also has a faster learning potential to find the optimized path to the goal. Additionally, this method also has a good ability to complete the goal-directed navigation task in large space and in the environments with obstacles.

• Smart Structures and Systems
• /
• v.15 no.1
• /
• pp.99-117
• /
• 2015

The Stonecutters Bridge (SCB) in Hong Kong is the third-longest cable-stayed bridge in the world with a main span stretching 1,018 m between two 298 m high single-leg tapering composite towers. A Wind and Structural Health Monitoring System (WASHMS) is being implemented on SCB by the Highways Department of The Hong Kong SAR Government, and the SCB-WASHMS is composed of more than 1,300 sensors in 15 types. In order to establish a linkage between structural health monitoring and maintenance management, a Structural Health Rating System (SHRS) with relevant rating tools and indices is devised. On the basis of a 3D space frame finite element model (FEM) of SCB and model updating, this paper presents the development of an SHR-oriented 3D multi-scale FEM for the purpose of load-resistance analysis and damage evaluation in structural element level, including modeling, refinement and validation of the multi-scale FEM. The refined 3D structural segments at deck and towers are established in critical segment positions corresponding to maximum cable forces. The components in the critical segment region are modeled as a full 3D FEM and fitted into the 3D space frame FEM. The boundary conditions between beam and shell elements are performed conforming to equivalent stiffness, effective mass and compatibility of deformation. The 3D multi-scale FEM is verified by the in-situ measured dynamic characteristics and static response. A good agreement between the FEM and measurement results indicates that the 3D multi-scale FEM is precise and efficient for WASHMS and SHRS of SCB. In addition, stress distribution and concentration of the critical segments in the 3D multi-scale FEM under temperature loads, static wind loads and equivalent seismic loads are investigated. Stress concentration elements under equivalent seismic loads exist in the anchor zone in steel/concrete beam and the anchor plate edge in steel anchor box of the towers.

• Smart Structures and Systems
• /
• v.14 no.3
• /
• pp.285-305
• /
• 2014

This study employs a novel beam-type wavelet finite element model (WFEM) to fulfill an adaptive-scale damage detection strategy in which structural modeling scales are not only spatially varying but also dynamically changed according to actual needs. Dynamical equations of beam structures are derived in the context of WFEM by using the second-generation cubic Hermite multiwavelets as interpolation functions. Based on the concept of modal strain energy, damage in beam structures can be detected in a progressive manner: the suspected region is first identified using a low-scale structural model and the more accurate location and severity of the damage can be estimated using a multi-scale model with local refinement in the suspected region. Although this strategy can be implemented using traditional finite element methods, the multi-scale and localization properties of the WFEM considerably facilitate the adaptive change of modeling scales in a multi-stage process. The numerical examples in this study clearly demonstrate that the proposed damage detection strategy can progressively and efficiently locate and quantify damage with minimal computation effort and a limited number of sensors.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.8
• /
• pp.2787-2800
• /
• 2022

The aging society increases emergency situations of the elderly living alone and a variety of social crimes. In order to prevent them, techniques to detect emergency situations through voice are actively researched. This study proposes CutPaste-based anomaly detection model using multi-scale feature extraction in time series streaming data. In the proposed method, an audio file is converted into a spectrogram. In this way, it is possible to use an algorithm for image data, such as CNN. After that, mutli-scale feature extraction is applied. Three images drawn from Adaptive Pooling layer that has different-sized kernels are merged. In consideration of various types of anomaly, including point anomaly, contextual anomaly, and collective anomaly, the limitations of a conventional anomaly model are improved. Finally, CutPaste-based anomaly detection is conducted. Since the model is trained through self-supervised learning, it is possible to detect a diversity of emergency situations as anomaly without labeling. Therefore, the proposed model overcomes the limitations of a conventional model that classifies only labelled emergency situations. Also, the proposed model is evaluated to have better performance than a conventional anomaly detection model.

• Steel and Composite Structures
• /
• v.36 no.1
• /
• pp.87-101
• /
• 2020

A multi-scale epoxy/CNT/fiberglass annular sector plate is studied in this paper in the view of determining nonlinear forced vibration characteristics. A 3D Mori-Tanaka model is employed for evaluating multi-scale material properties. Thus, all of glass fibers are assumed to have uni-direction alignment and CNTs have random diffusion. The geometry of annular sector plate can be described based on the open angle and the value of inner/outer radius. In order to solve governing equations and derive exact forced vibration curves for the multi-scale annular sector, Jacobi elliptic functions are used. Obtained results demonstrate the significance of CNT distribution, geometric nonlinearity, applied force, fiberglass volume, open angle and fiber directions on forced vibration characteristics of multi-scale annular sector plates.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.4
• /
• pp.980-997
• /
• 2024

To improve the brightness of images and reveal hidden information in dark areas is the main objective of low-light image enhancement (LLIE). LLIE methods based on deep learning show good performance. However, there are some limitations to these methods, such as the complex network model requires highly configurable environments, and deficient enhancement of edge details leads to blurring of the target content. Single-scale feature extraction results in the insufficient recovery of the hidden content of the enhanced images. This paper proposed an edge detection-based multi-scale feature enhancement network for LLIE (EDMFEN). To reduce the loss of edge details in the enhanced images, an edge extraction module consisting of a Sobel operator is introduced to obtain edge information by computing gradients of images. In addition, a multi-scale feature enhancement module (MSFEM) consisting of multi-scale feature extraction block (MSFEB) and a spatial attention mechanism is proposed to thoroughly recover the hidden content of the enhanced images and obtain richer features. Since the fused features may contain some useless information, the MSFEB is introduced so as to obtain the image features with different perceptual fields. To use the multi-scale features more effectively, a spatial attention mechanism module is used to retain the key features and improve the model performance after fusing multi-scale features. Experimental results on two datasets and five baseline datasets show that EDMFEN has good performance when compared with the stateof-the-art LLIE methods.

• 한국전산유체공학회:학술대회논문집
• /
• 2006.05a
• /
• pp.368-371
• /
• 2006

• Transactions of Materials Processing
• /
• v.21 no.6
• /
• pp.366-371
• /
• 2012

Atomistic simulations have become useful tools for exploring new insights in materials science, but the length and time scale that can be handled with atomistic simulations are seriously limiting their practical applications. In order to make meaningful quantitative predictions, atomistic simulations are necessarily combined with higher-scale modeling. The present research is thus concerned with the development of a multi-scale model and its application to the prediction of the mechanical properties of body-centered cubic(BCC) iron with an emphasis on the coupling of atomistic molecular dynamics with meso-scale discrete dislocation dynamics modeling. In order to achieve predictive multi-scale simulations, it is necessary to properly incorporate atomistic details into the meso-scale approach. This challenge is handled with the proposed hierarchical information passing strategy from atomistic to meso-scale by obtaining material properties and dislocation mobility. Finally, this fundamental and physics-based meso-scale approach is employed for quantitative predictions of the mechanical response of single crystal iron.

• Advances in biomechanics and applications
• /
• v.1 no.1
• /
• pp.23-40
• /
• 2014

This paper examines the blood chamber of a left ventricular assist device (LVAD) under static loading conditions and standard operating temperatures. The LVAD's walls are made of a temperature-sensitive polymer (ChronoFlex C 55D) and are covered with a titanium nitride (TiN) nano-coating (deposited by laser ablation) to improve their haemocompatibility. A loss of cohesion may be observed near the coating-substrate boundary. Therefore, a micro-scale stress-strain analysis of the multilayered blood chamber was conducted with FE (finite element) code. The multi-scale model included a macro-model of the LVAD's blood chamber and a micro-model of the TiN coating. The theories of non-linear elasticity and elasto-plasticity were applied. The formulated problems were solved with a finite element method. The micro-scale problem was solved for a representative volume element (RVE). This micro-model accounted for the residual stress, a material model of the TiN coating, the stress results under loading pressures, the thickness of the TiN coating and the wave parameters of the TiN surface. The numerical results (displacements and strains) were experimentally validated using digital image correlation (DIC) during static blood pressure deformations. The maximum strain and stress were determined at static pressure steps in a macro-scale FE simulation. The strain and stress were also computed at the same loading conditions in a micro-scale FE simulation.