• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.11a
• /
• pp.44-45
• /
• 2019

최근 CNN 등 인공신경망은 최근 이미지 분류, 객체 인식, 자연어 처리 등 다양한 분야에서 뛰어난 성능을 보이고 있다. 그러나, 대부분의 분야에서 보다 더 높은 성능을 얻기 위해 사용한 인공신경망 모델들은 파라미터 수 및 연산량 등이 방대하여, 모바일 및 IoT 디바이스 같은 연산량이나 메모리가 제한된 환경에서 추론하기에는 제한적이다. 따라서 연산량 및 모델 파라미터 수를 압축하기 위한 딥러닝 경량화 알고리즘이 연구되고 있다. 본 논문에서는 임베디트 보드에서의 압축된 CNN 모델의 성능을 검증한다. 인공지능 지원 맞춤형 칩인 QCS605 를 내장한 임베디드 보드에서 카메라로 입력한 영상에 대해서 원 CNN 모델과 압축된 CNN 모델의 분류 성능과 동작속도 비교 분석한다. 본 논문의 실험에서는 CNN 모델로 MobileNetV2, VGG16 을 사용했으며, 주어진 모델에서 가지치기(pruning) 기법, 양자화, 행렬 분해 등의 인공신경망 압축 기술을 적용하였을 때 원래의 모델 대비 추론 시간 및 분류의 정확도 성능을 분석하고 인공신경망 압축 기술의 유용성을 확인하였다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.17 no.3
• /
• pp.310-315
• /
• 2007

When a mail was given to users, each user's response could be different according to his or her preference. This paper presents a solution for this situation by constructing a u!;or preferred ontology for anti-spam systems. To define an ontology for describing user behaviors, we applied associative classification mining to study preference information of users and their responses to emails. Generated classification rules can be represented in a formal ontology language. A user preferred ontology can explain why mail is decided to be spam or non-spam in a meaningful way. We also suggest a nor rule optimization procedure inspired from logic synthesis to improve comprehensibility and exclude redundant rules.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.43 no.12 s.354
• /
• pp.42-52
• /
• 2006

The interest for IPv6, Which is next generation IP, increases, while th difficulty for adopting it into IT indestries also increase too. We also expect that Internet technologies develop 0 basis of All-IP networks, which include various heterogeneous networks. Therefore, it is necessary for network to support mobility and multicast capavility to create application with wide-variety- In this paper, we propose a scheme that be able to provide mobility and multicast efficient in IPv6 networks. The proposed scheme has been implemented and experimented on NS-2 and validated with respect to multicast packet delay time and count of Joining and pruning)

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.12
• /
• pp.3219-3236
• /
• 2012

A tree routing structure is often adopted for many-to-one data gathering and aggregation in sensor networks. For real-time scenarios, considering lossy wireless links, it is an important issue how to construct a maximum-lifetime data gathering tree with delay constraint. In this work, we study the problem of lifetime-preserving and delay-constrained tree construction in unreliable sensor networks. We prove that the problem is NP-complete. A greedy approximation algorithm is proposed. We use expected transmissions count (ETX) as the link quality indicator, as well as a measure of delay. Our algorithm starts from an arbitrary least ETX tree, and iteratively adjusts the hierarchy of the tree to reduce the load on bottleneck nodes by pruning and grafting its sub-tree. The complexity of the proposed algorithm is $O(N^4)$. Finally, extensive simulations are carried out to verify our approach. Simulation results show that our algorithm provides longer lifetime in various situations compared to existing data gathering schemes.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.141-151
• /
• 2022

In recent years, the industry and research communities have focused on developing autonomous crack inspection approaches, which mainly include image acquisition and crack detection. In these approaches, mobile devices such as cameras, drones or smartphones are utilized as sensing platforms to acquire structural images, and the deep learning (DL)-based methods are being developed as important crack detection approaches. However, the process of image acquisition and collection is time-consuming, which delays the inspection. Also, the present mobile devices such as smartphones can be not only a sensing platform but also a computing platform that can be embedded with deep neural networks (DNNs) to conduct on-site crack detection. Due to the limited computing resources of mobile devices, the size of the DNNs should be reduced to improve the computational efficiency. In this study, an architecture called pruned crack recognition network (PCR-Net) was developed for the detection of structural cracks. A dataset containing 11000 images was established based on the raw images from bridge inspections. A pruning method was introduced to reduce the size of the base architecture for the optimization of the model size. Comparative studies were conducted with image processing techniques (IPTs) and other DNNs for the evaluation of the performance of the proposed PCR-Net. Furthermore, a modularly designed framework that integrated the PCR-Net was developed to realize a DL-based crack detection application for smartphones. Finally, on-site crack detection experiments were carried out to validate the performance of the developed system of smartphone-based detection of structural cracks.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.5
• /
• pp.169-176
• /
• 2020

Recently, deep neural networks (DNNs) have brought revolutions to many mobile and embedded devices by providing human-level machine intelligence for various applications. However, high inference accuracy of such DNNs comes at high computational costs, and, hence, there have been significant efforts to reduce computational overheads of DNNs either by compressing off-the-shelf models or by designing a new small footprint DNN architecture tailored to resource constrained devices. One notable recent paradigm in designing small footprint DNN models is sharing parameters in several layers. However, in previous approaches, the parameter-sharing techniques have been applied to large deep networks, such as ResNet, that are known to have high redundancy. In this paper, we propose a parameter-sharing method for already parameter-efficient small networks such as ShuffleNetV2. In our approach, small templates are combined with small layer-specific parameters to generate weights. Our experiment results on ImageNet and CIFAR100 datasets show that our approach can reduce the size of parameters by 15%-35% of ShuffleNetV2 while achieving smaller drops in accuracies compared to previous parameter-sharing and pruning approaches. We further show that the proposed approach is efficient in terms of latency and energy consumption on modern embedded devices.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.10
• /
• pp.79-89
• /
• 2019

While most research focuses on the k-nearest neighbors (kNN) queries in the database community, an important type of proximity queries called k-farthest neighbors (kFN) queries has not received much attention. This paper addresses the problem of finding the k-farthest neighbors in road networks. Given a positive integer k, a query object q, and a set of data points P, a kFN query returns k data objects farthest from the query object q. Little attention has been paid to processing kFN queries in road networks. The challenge of processing kFN queries in road networks is reducing the number of network distance computations, which is the most prominent difference between a road network and a Euclidean space. In this study, we propose an efficient algorithm called FANS for k-FArthest Neighbor Search in road networks. We present a shared computation strategy to avoid redundant computation of the distances between a query object and data objects. We also present effective pruning techniques based on the maximum distance from a query object to data segments. Finally, we demonstrate the efficiency and scalability of our proposed solution with extensive experiments using real-world roadmaps.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.11
• /
• pp.2367-2372
• /
• 2009

Recently, Many studies have been progressing to path-finding adapted dynamic information on the Terminal Based Navigation System(TBNS). The algorithms proposed are based on $A^*$ algorithm. Path-finding algorithms which use heuristic function may occur a problem of the increase of exploring cost. Path-finding with an abstract graph which expresses real road network as a simple graph is proposed for reducing dependency of heuristic and exploring cost. In this paper, two abstract graph that are different method of construction, Homogeneous Node merging($AG^H$) and Connected Node Merging($AG^C$), are implemented. In result of evaluation of performance, $AG^C$ has better performance than $AG^H$ at construction cost and the number of node access but $AG^C$ has worse performance than AGH at exploring cost.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.10a
• /
• pp.245-248
• /
• 2009

Recently, Many studies have been progressing to path-finding with dynamic information on the Terminal Based Navigation System(TBNS). However, the most of existing algorithms are based on $A{\ast}$ algorithm. Path-finding algorithms which use heuristic function may occur a problem of the increase of exploring cost in case of that there is no way determined by heuristic function or there are 2 way more which have almost same cost. In this paper, two abstract graph(AG) that are different method of construction, Homogeneous Node merging($AG^H$) and Connected Node Merging($AG^C$), are implemented. The abstract graph is a simple graph of real road network. The method of using the abstract graph is proposed for reducing dependency of heuristic and exploring cost. In result of evaluation of performance, $AG^C$ has better performance than $AG^H$ at construction cost but $AG^C$ has worse performance than $AG^H$ exploring cost.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.46 no.1
• /
• pp.55-67
• /
• 2023

Lightweight face recognition models, as one of the most popular and long-standing topics in the field of computer vision, has achieved vigorous development and has been widely used in many real-world applications due to fewer number of parameters, lower floating-point operations, and smaller model size. However, few surveys reviewed lightweight models and reimplemented these lightweight models by using the same calculating resource and training dataset. In this survey article, we present a comprehensive review about the recent research advances on the end-to-end efficient lightweight face recognition models and reimplement several of the most popular models. To start with, we introduce the overview of face recognition with lightweight models. Then, based on the construction of models, we categorize the lightweight models into: (1) artificially designing lightweight FR models, (2) pruned models to face recognition, (3) efficient automatic neural network architecture design based on neural architecture searching, (4) Knowledge distillation and (5) low-rank decomposition. As an example, we also introduce the SqueezeFaceNet and EfficientFaceNet by pruning SqueezeNet and EfficientNet. Additionally, we reimplement and present a detailed performance comparison of different lightweight models on the nine different test benchmarks. At last, the challenges and future works are provided. There are three main contributions in our survey: firstly, the categorized lightweight models can be conveniently identified so that we can explore new lightweight models for face recognition; secondly, the comprehensive performance comparisons are carried out so that ones can choose models when a state-of-the-art end-to-end face recognition system is deployed on mobile devices; thirdly, the challenges and future trends are stated to inspire our future works.