• 제어로봇시스템학회논문지
• /
• 제5권7호
• /
• pp.841-847
• /
• 1999

For successful assembly of flexible parts, informations about their deformation as well as possible misalignments between the holes and their mating parts are essential. Such informations can be acquired from visual sensors. For robotic assembly, the corrective assembly motion to compensate for such misalignments has to be determined from the measured informations. However, this may not be simply derived from the measured misalignment alone because the part deformation progressively occurs during misalignment compensation. Based on the analysis of flexible parts assembly process, this paper presents a neural net-based inference system that can infer the complex relationship between the corrective motion and the measured information of parts deformation and misalignments. And it verifies the performance of the implemented inference system. The results show that the proposed neural net-based misalignment compensation algorithm Is effective in compensating for the lateral misalignment, and that it can be extended to the assembly tasks under more general conditions.

• Geomechanics and Engineering
• /
• 제20권5호
• /
• pp.385-397
• /
• 2020

This paper presents an application of artificial neural networks (ANNs) in settlement prediction of a foundation on sandy soil. In order to train the ANN model, a wide experimental database about settlement of foundations acquired from available literatures was collected. The data used in the ANNs model were arranged using the following five-input parameters that covered both geometrical foundation and sandy soil properties: breadth of foundation B, length to width L/B, embedment ratio D_f/B, foundation net applied pressure q_net, and average SPT blow count N. The backpropagation algorithm was implemented to develop an explicit predicting formulation. The settlement results are compared with the results of previous studies. The accuracy of the proposed formula proves that the ANNs method has a huge potential for predicting the settlement of foundations on sandy soils.

• KIEE International Transaction on Systems and Control
• /
• 제2D권2호
• /
• pp.108-114
• /
• 2002

This paper describes the design of a robust adaptive controller for a nonlinear dynamical system with unknown nonlinearities. These unknown nonlinearities are approximated by multilayered neural networks (MNNs) whose parameters are adjusted on-line, according to some adaptive laws far controlling the output of the nonlinear system, to track a given trajectory. The main contribution of this paper is a method for considering input constraint with a rigorous stability proof. The Lyapunov synthesis approach is used to develop a state-feedback adaptive control algorithm based on the adaptive MNN model. An overall control system guarantees that the tracking error converges at about zero and that all signals involved are uniformly bounded even in the presence of input saturation. Theoretical results are illustrated through a simulation example.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2022년도 춘계학술대회
• /
• pp.636-638
• /
• 2022

본 논문은 의료진단 검출 분야에서 혈관, 신경조직, 망막 손상 그리고 다양한 심혈관계 질환과 치매까지 진단하는 데 유용하게 사용하고 있는 안저 영상에 CNN(Convolution Neural Network) 알고리즘을 적용하고 녹내장 병변을 검출하기 위한 연구를 진행한다. 실험을 위하여 정상 안저 영상과 녹내장 병변이 있는 안저 영상으로 구성된 데이터 세트를 AlexNet으로 분류하고 그 성능을 확인하였다.

• Communications for Statistical Applications and Methods
• /
• 제29권2호
• /
• pp.161-176
• /
• 2022

There has been active research in image classification using deep learning convolutional neural network (CNN) models. ImageNet large-scale visual recognition challenge (ILSVRC) (2010-2017) was one of the most important competitions that boosted the development of efficient deep learning algorithms. This paper introduces and compares six monumental models that achieved high prediction accuracy in ILSVRC. First, we provide a review of the models to illustrate their unique structure and characteristics of the models. We then compare those models under a unified framework. For this reason, additional devices that are not crucial to the structure are excluded. Four popular data sets with different characteristics are then considered to measure the prediction accuracy. By investigating the characteristics of the data sets and the models being compared, we provide some insight into the architectural features of the models.

• ETRI Journal
• /
• 제46권2호
• /
• pp.307-322
• /
• 2024

Most object detection methods use a horizontal bounding box that causes problems between adjacent objects with arbitrary directions, resulting in misaligned detection. Hence, the horizontal anchor should be replaced by a rotating anchor to determine oriented bounding boxes. A two-stage process of delineating a horizontal bounding box and then converting it into an oriented bounding box is inefficient. To improve detection, a box-boundary-aware vector can be estimated based on a convolutional neural network. Specifically, we propose a ResNeXt101 encoder to overcome the weaknesses of the conventional ResNet, which is less effective as the network depth and complexity increase. Owing to the cardinality of using a homogeneous design and multi-branch architecture with few hyperparameters, ResNeXt captures better information than ResNet. Experimental results demonstrate more accurate and faster oriented object detection of our proposal compared with a baseline, achieving a mean average precision of 89.41% and inference rate of 23.67 fps.

• Korean Journal of Radiology
• /
• 제22권3호
• /
• pp.476-488
• /
• 2021

Objective: We aimed to develop a deep neural network for segmenting lung parenchyma with extensive pathological conditions on non-contrast chest computed tomography (CT) images. Materials and Methods: Thin-section non-contrast chest CT images from 203 patients (115 males, 88 females; age range, 31-89 years) between January 2017 and May 2017 were included in the study, of which 150 cases had extensive lung parenchymal disease involving more than 40% of the parenchymal area. Parenchymal diseases included interstitial lung disease (ILD), emphysema, nontuberculous mycobacterial lung disease, tuberculous destroyed lung, pneumonia, lung cancer, and other diseases. Five experienced radiologists manually drew the margin of the lungs, slice by slice, on CT images. The dataset used to develop the network consisted of 157 cases for training, 20 cases for development, and 26 cases for internal validation. Two-dimensional (2D) U-Net and three-dimensional (3D) U-Net models were used for the task. The network was trained to segment the lung parenchyma as a whole and segment the right and left lung separately. The University Hospitals of Geneva ILD dataset, which contained high-resolution CT images of ILD, was used for external validation. Results: The Dice similarity coefficients for internal validation were 99.6 ± 0.3% (2D U-Net whole lung model), 99.5 ± 0.3% (2D U-Net separate lung model), 99.4 ± 0.5% (3D U-Net whole lung model), and 99.4 ± 0.5% (3D U-Net separate lung model). The Dice similarity coefficients for the external validation dataset were 98.4 ± 1.0% (2D U-Net whole lung model) and 98.4 ± 1.0% (2D U-Net separate lung model). In 31 cases, where the extent of ILD was larger than 75% of the lung parenchymal area, the Dice similarity coefficients were 97.9 ± 1.3% (2D U-Net whole lung model) and 98.0 ± 1.2% (2D U-Net separate lung model). Conclusion: The deep neural network achieved excellent performance in automatically delineating the boundaries of lung parenchyma with extensive pathological conditions on non-contrast chest CT images.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2001년도 추계학술대회 학술발표 논문집
• /
• pp.10-13
• /
• 2001

Nonlinear Function Approximation of Moduled Neural Network Using Genetic Algorithm Neural Network consists of neuron and synapse. Synapse memorize last pattern and study new pattern. When Neural Network learn new pattern, it tend to forget previously learned pattern. This phenomenon is called to catastrophic inference or catastrophic forgetting. To overcome this phenomenon, Neural Network must be modularized. In this paper, we propose Moduled Neural Network. Modular Neural Network consists of two Neural Network. Each Network individually study different pattern and their outputs is finally summed by net function. Sometimes Neural Network don't find global minimum, but find local minimum. To find global minimum we use Genetic Algorithm.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제8권5호
• /
• pp.205-212
• /
• 2019

장면 그래프는 영상 내 물체들과 각 물체 간의 관계를 나타내는 지식 그래프를 의미한다. 본 논문에서는 3차원 실내 환경을 위한 3차원 장면 그래프를 생성하는 모델을 제안한다. 3차원 장면 그래프는 물체들의 종류와 위치, 그리고 속성들뿐만 아니라, 물체들 간의 3차원 공간 관계들도 포함한다. 따라서 3차원 장면 그래프는 에이전트가 활동할 실내 환경을 묘사하는 하나의 사전 지식 베이스로 볼 수 있다. 이러한 3차원 장면 그래프는 영상 기반의 질문과 응답, 서비스 로봇 등과 같은 다양한 분야에서 유용하게 활용될 수 있다. 본 논문에서 제안하는 3차원 장면 그래프 생성 모델은 크게 물체 탐지 네트워크(ObjNet), 속성 예측 네트워크(AttNet), 변환 네트워크(TransNet), 관계 예측 네트워크(RelNet) 등 총 4가지 부분 네트워크들로 구성된다. AI2-THOR가 제공하는 3차원 실내 가상환경들을 이용한 다양한 실험들을 통해, 본 논문에서 제안한 모델의 높은 성능을 확인할 수 있었다.

• 한국전자통신학회논문지
• /
• 제17권3호
• /
• pp.529-536
• /
• 2022

지난 2년 동안 중증급성호흡기증후군 코로나바이러스-2(SARS-CoV-2)는 점점 더 많은 사람들에게 영향을 미치고 있다. 본 논문에서는 COVID-19 폐 CT 이미지를 분할하고 분류하기 위해서 서브코딩블록(SCB), 확장공간파라미드풀링(ASSP)와 어텐션게이트(AG)로 구성된 혼합 모드 특징 추출 방식의 새로운 U-Net 컨볼루션 신경망을 제안한다. 그리고 제안된 모델과 비교하기 위하여 FCN, U-Net, U-Net-SCB 모델을 설계한다. 제안된 U-Net-MMFE 는 COVID-19 CT 스캔 디지털 이미지 데이터에 대하여 atrous rate가 12이고, Adam 최적화 알고리즘을 사용할 때 다른 분할 모델에 비하여 94.79%의 우수한 주사위 분할 점수를 얻었다.