• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.4
• /
• pp.381-389
• /
• 2022

Detecting partially occluded objects is difficult due to the appearances and shapes of occluders are highly variable. These variabilities lead to challenges of localizing accurate bounding box or classifying objects with visible object parts. To address these problems, we propose a two-stage part-based attention approach for robust object detection under partial occlusion. First, our part attention network(PAN) captures the important object parts and then it is used to generate weighted object features. Based on the weighted features, the re-weighted object features are produced by our reinforced PAN(RPAN). Experiments are performed on our collected military vehicle dataset and synthetic occlusion dataset. Our method outperforms the baselines and demonstrates the robustness of detecting objects under partial occlusion.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2008.10a
• /
• pp.329-341
• /
• 2008

Knowledge management has garnered attention due to its role of maintaining competitive advantage. Creating and sharing knowledge is an essential part of managing knowledge. However, the best knowledge is underutilized because employees tend to seek knowledge through their informal networks, not reach out to other sources for obtaining the best knowledge. Prior studies on informal networks pointed out a negative influence of heavy reliance on learning through informal networks but they paid little attention to a structure of informal networks and its impacts on diffusion of knowledge. The aim of our study is to show impacts of informal network on knowledge management by employing a network structure and investigating diffusion of knowledge within it. Our study found out that performance of learning becomes lower in a highly clustered network. Creating random links such as serendipitous learning can improve performance of knowledge management. When employees rely on a knowledge management system, creating random links is not necessary. Costs of adopting knowledge affect performance of knowledge management.

• Journal of Information Processing Systems
• /
• v.19 no.1
• /
• pp.98-108
• /
• 2023

This paper is an attempt to design segmentation method based on fully convolutional networks (FCN) and attention mechanism. The first five layers of the Visual Geometry Group (VGG) 16 network serve as the coding part in the semantic segmentation network structure with the convolutional layer used to replace pooling to reduce loss of image feature extraction information. The up-sampling and deconvolution unit of the FCN is then used as the decoding part in the semantic segmentation network. In the deconvolution process, the skip structure is used to fuse different levels of information and the attention mechanism is incorporated to reduce accuracy loss. Finally, the segmentation results are obtained through pixel layer classification. The results show that our method outperforms the comparison methods in mean pixel accuracy (MPA) and mean intersection over union (MIOU).

• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.4
• /
• pp.145-151
• /
• 2021

In this paper, we propose a novel visualization technique to explain the predictions of deep neural networks. We use knowledge distillation (KD) to identify the interior of a black-box model for which we know only inputs and outputs. The information of the black box model will be transferred to a white box model that we aim to create through the KD. The white box model will learn the representation of the black-box model. Second, the white-box model generates attention maps for each of its layers using Grad-CAM. Then we combine the attention maps of different layers using the pixel-wise summation to generate a final saliency map that contains information from all layers of the model. The experiments show that the proposed technique found important layers and explained which part of the input is important. Saliency maps generated by the proposed technique performed better than those of Grad-CAM in deletion game.

• The KIPS Transactions:PartC
• /
• v.18C no.1
• /
• pp.45-50
• /
• 2011

Due to advances in wireless communications and portable terminals, any-time, any-where, and any-device Internet access is possible. In particular, Internet access in moving vehicles is an emerging and challenging issue. Even though a variety studies have been conduced for vehicular networks, little attention is paid to vehicular Wi-Fi networks where a Wi-Fi access point (AP) is installed at the vehicle and the AP is connected to an external base station (BS). In this paper, we conduct a measurement study on the uplink and downlink throughput for Internet access in vehicular Wi-Fi networks. We consider diverse network environments: high-speed train, car, and subway. Measurement results demonstrate that current Internet access in vehicular Wi-Fi networks are not satisfactory for interactive multimedia applications. Therefore, in-depth study on resource management in vehicular Wi-Fi networks is strongly required.

• The KIPS Transactions:PartC
• /
• v.16C no.4
• /
• pp.521-526
• /
• 2009

Disruption/Delay Tolerant Network(DTN) is a technology for interconnecting partitioned networks. These days, DTN, especially routing in DTN, draws significant attention from the networking community. In this paper, we investigate DTN routing strategies for highly partitioned ad hoc networks where Unmanned Aerial Vehicles (UAVs) perform store-carry-forward functionality for improved network connectivity. Also we investigate UAV trajectory control mechanisms via simulation studies.

• The KIPS Transactions:PartC
• /
• v.13C no.4 s.107
• /
• pp.441-446
• /
• 2006

In recent years, WDM networks have received much attention as the Internet backbone networks because of the explosive growth of the Internet IP-based traffic. The Optical Burst Switching (OBS) has been proposed as an effective optical switching technology in the WDM networks. The OBS has the advantages in 1) the high usage rate of the bandwidth, and 2) no necessity of optical buffer. However, the OBS has the burst-contention problem in the networks. The deflection routing is proposed as one of means to solve this problem. In this paper, we propose a new routing method to minimize burst loss in the deflection routing based networks. In addition, we propose a QoS control method using a new routing algorithm. Finally, we show the variety of the proposed methods by computer simulations.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.10
• /
• pp.3685-3707
• /
• 2021

In recent years, facial attribute editing has been successfully used to effectively change face images of various attributes based on generative adversarial networks and encoder-decoder models. However, existing models have a limitation in that they may change an unintended part in the process of changing an attribute or may generate an unnatural result. In this paper, we propose a model that improves the learning of the attention mask by adding a spatial attention mechanism based on the unified selective transfer network (referred to as STGAN) using semi-supervised learning. The proposed model can edit multiple attributes while preserving details independent of the attributes being edited. This study makes two main contributions to the literature. First, we propose an encoder-decoder model structure that learns and edits multiple facial attributes and suppresses distortion using an attention mask. Second, we define guide masks and propose a method and an objective function that use the guide masks for multiple facial attribute editing through semi-supervised learning. Through qualitative and quantitative evaluations of the experimental results, the proposed method was proven to yield improved results that preserve the image details by suppressing unintended changes than existing methods.

• Journal of Advanced Navigation Technology
• /
• v.15 no.1
• /
• pp.54-60
• /
• 2011

The development of wireless sensor networks (WSN) can be motivated by several types of applications. However, these applications demand an energy-efficient WSN that can prolong the network lifetime and can provide high throughput, low latency and delay. Designing wireless sensor networks with the capability of prolonging network lifetime catch the attention of many researchers in wireless system and network field. Contrasts with Mobile Ad Hoc Network system, Wireless Sensor Networks designs focused more on survivability of each node in the network instead of maximizing data throughput or minimizing end-to-end delay. In this paper, we will study part of data link layer in Open Systems Interconnection (OSI) model, called medium access control (MAC) layer. Since the MAC development of energy aware MAC Protocol for wireless sensor layer controls the physical radio part, it has a large impact on the overall energy consumption and the lifetime of a node. This paper proposes a analytical approach that tries to reduce idle energy consumption, and shows the increasement of network end-to-end arrival rate due to efficiency in energy consumption from time slot management.

• The KIPS Transactions:PartC
• /
• v.18C no.6
• /
• pp.423-432
• /
• 2011

Recently, OpenFlow has been paid attention to as a fundamental technology which provides a function of virtualization and programmability in network. In Korea, deployment of OpenFlow networks in campuses and the interconnection between them through tunneling in layer 3 has been performed. However, the performance of the interconnected networks is decreased due to delay in IP layer. In this paper, we design and deploy nation-wide, not local, OpenFlow networks in a pure layer 2 environment over KREONET. After that, we do end-to-end Round-trip Time measurements and TCP/UDP performance tests in OpenFlow and normal networks, and do comparison and analysis on the test results. The results show that the nation-wide OpenFlow networks provide equal performance to normal networks except for the initial packet loss for UDP streaming. In regards to the performance decrease due to early UDP packet loss, we can mitigate it by implementing exceptional procedures in a controller which deal with the same continuous "Packet_in" events.