• Smart Media Journal
• /
• v.7 no.4
• /
• pp.61-69
• /
• 2018

Multi-tracking of general objects and specific faces is an important topic in the field of computer vision applicable to many branches of industry such as biometrics, security, etc. The rapid development of deep neural networks has resulted in a dramatic improvement in face recognition and object detection problems, which helps improve the multiple-face tracking techniques exploiting the tracking-by-detection method. Our proposed method uses face detection trained with a head dataset to resolve the face deformation problem in the tracking process. Further, we use robust face features extracted from the deep face recognition network to match the tracklets with tracking faces using Hungarian matching method. We achieved promising results regarding the usage of deep face features and head detection in a face tracking benchmark.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.244-247
• /
• 2020

Data augmentation has been helpful in improving the performance in deep learning, when we have a limited data and random erasing is one of the augmentations that have shown impressive performance in deep learning in multiple domains. But the main issue is that sometime it loses good features when randomly selected region is erased by some random values, that does not improve performance as it should. We target that problem in way that good features should not be lost and also want random erasing at the same time. For that purpose, we introduce new augmentation technique named Intra-Class Random Erasing (ICRE) that focuses on data to learn robust features of the same class samples by randomly exchanging randomly selected region. We perform multiple experiments by using different models including resnet18, VGG16 over variety of the datasets including ESC10, UrbanSound8K. Our approach has shown effectiveness over others methods including random erasing.

• Journal of The Korean Astronomical Society
• /
• v.41 no.6
• /
• pp.181-186
• /
• 2008

Recently, Choe & Cheng (2002) have demonstrated that multiple magnetic flux systems with closed configurations can have more magnetic energy than the corresponding open magnetic fields. In relation to this issue, we have addressed two questions: (1) how much fraction of eruptive solar active regions shows multiple flux system features, and (2) what winding angle could be an eruption threshold. For this investigation, we have taken a sample of 105 front-side halo CMEs, which occurred from 1996 to 2001, and whose source regions were located near the disk center, for which magnetic polarities in SOHO/MDI magnetograms are clearly discernible. Examining their soft X-ray images taken by Yohkoh SXT in pre-eruption stages, we have classified these events into two groups: multiple flux system events and single flux system events. It is found that 74% (78/105) of the sample events show multiple flux system features. Comparing the field configuration of an active region with a numerical model, we have also found that the winding angle of the eruptive flux system is slightly above $1.5{\pi}$.

• Journal of Broadcast Engineering
• /
• v.17 no.4
• /
• pp.634-639
• /
• 2012

In this paper, we propose nonlinear optimization method for feature matching from multiple view image. Typical solution of feature matching is by solving linear equation. However this solution has large error due to nonlinearity of image formation model. If typical nonlinear optimization method is used, complexity grows exponentially over the number of features. To make complexity lower, we use sparse Levenberg-Marquardt nonlinear optimization for matching of features over multiple view image.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.36 no.3
• /
• pp.17-24
• /
• 2013

Release planning in a software product line (SPL) is to select and assign the features of the multiple software products in the SPL in sequence of releases along a specified planning horizon satisfying the numerous constraints regarding technical precedence, conflicting priorities for features, and available resources. A greedy genetic algorithm is designed to solve the problems of release planning in SPL which is formulated as a precedence-constrained multiple 0-1 knapsack problem. To be guaranteed to obtain feasible solutions after the crossover and mutation operation, a greedy-like heuristic is developed as a repair operator and reflected into the genetic algorithm. The performance of the proposed solution methodology in this research is tested using a fractional factorial experimental design as well as compared with the performance of a genetic algorithm developed for the software release planning. The comparison shows that the solution approach proposed in this research yields better result than the genetic algorithm.

• Smart Structures and Systems
• /
• v.23 no.4
• /
• pp.393-403
• /
• 2019

Applying more features gives us better accuracy in modeling; however, increasing the inputs causes the curse of dimensions. In this paper, a new structure has been proposed for fault detecting and identifying (FDI) of high-dimensional systems. This structure consist of two structure. The first part includes Auto-Encoders (AE) as Deep Neural Networks (DNNs) to produce feature engineering process and summarize the features. The second part consists of the Local Model Networks (LMNs) with LOcally LInear MOdel Tree (LOLIMOT) algorithm to model outputs (multiple models). The fault detection is based on these multiple models. Hence the residuals generated by comparing the system output and multiple models have been used to alarm the faults. To show the effectiveness of the proposed structure, it is tested on single-shaft industrial gas turbine prototype model. Finally, a brief comparison between the simulated results and several related works is presented and the well performance of the proposed structure has been illustrated.

• International Journal of Internet, Broadcasting and Communication
• /
• v.13 no.1
• /
• pp.180-186
• /
• 2021

In cloud computing service it is important to share the system resource among multiple instances according to user requirements. In particular, the issue of efficiently distributing I/O resources across multiple instances is paid attention due to the rise of emerging data-centric technologies such as big data and deep learning. However, it is difficult to evaluate the I/O resource distribution of a Linux container, which is one of the core technologies of cloud computing, since conventional I/O benchmarks does not support features related to container management. In this paper, we propose a new I/O performance benchmarking framework that can easily evaluate the resource distribution of Linux containers using existing I/O benchmarks by supporting container-related features and integrated user interface. According to the performance evaluation result with trace-replay benchmark, the proposed benchmark framework has induced negligible performance overhead while providing convenience in evaluating the I/O performance of multiple Linux containers.

• The Journal of Korea Robotics Society
• /
• v.17 no.3
• /
• pp.353-358
• /
• 2022

Due to the limited field of view of the pinhole camera, there is a lack of stability and accuracy in camera pose estimation applications such as visual SLAM. Nowadays, multiple-camera setups and large field of cameras are used to solve such issues. However, a multiple-camera system increases the computation complexity of the algorithm. Therefore, in multiple camera-assisted visual simultaneous localization and mapping (vSLAM) the multi-view tracking algorithm is proposed that can be used to balance the budget of the features in tracking and local mapping. The proposed algorithm is based on PanoSLAM architecture with a panoramic camera model. To avoid the scale issue 3D LiDAR is fused with omnidirectional camera setup. The depth is directly estimated from 3D LiDAR and the remaining features are triangulated from pose information. To validate the method, we collected a dataset from the outdoor environment and performed extensive experiments. The accuracy was measured by the absolute trajectory error which shows comparable robustness in various environments.

• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.522-525
• /
• 2009

This paper proposed the content-based method of classifying filthy Phornographic video, which causes a big problem of modern society as the reverse function of internet. Audio data was used to extract the features from Phornographic video. There are frequency spectrum, autocorrelation, and MFCC as the feature of audio used in this paper. The sound that could be filthy contents was extracted, and the Phornographic was classified by measuring how much percentage of relevant sound was corresponding with the whole audio of video. For the experiment on the proposed method, The efficiency of classifying Phornographic was measured on each feature, and the measured result and comparison with using multi features were performed. I can obtain the better result than when only one feature of audio was extracted, and used.

• Journal of Korea Multimedia Society
• /
• v.22 no.12
• /
• pp.1457-1465
• /
• 2019

3D convolution is to stack multiple consecutive frames to form a cube, and then apply the 3D convolution kernel in the cube. In this structure, each feature map of the convolutional layer is connected to multiple adjacent sequential frames in the previous layer, thus capturing the motion information. However, due to the changes of pedestrian posture, motion and position, the convolution at the same place is inappropriate, and when the 3D convolution kernel is convoluted in the time domain, only time domain features of three consecutive frames can be extracted, which is not a good enough to get action information. This paper proposes an action recognition method based on feature fusion of 3D convolutional neural network. Based on the VGG16 network model, sending a pre-acquired optical flow image for learning, then get the time domain features, and then the feature of the time domain is extracted from the features extracted by the 3D convolutional neural network. Finally, the behavior classification is done by the SVM classifier.