• 인터넷정보학회논문지
• /
• 제19권3호
• /
• pp.35-42
• /
• 2018

딥 러닝은 영상 분류를 위한 여러 방법 중 높은 정확도를 보이는 방법으로 알려져 있다. 본 논문에서는 딥 러닝 방법 가운데 합성곱 신경망 (CNN:Convolutional Neural Network)을 이용하여 영상을 분류함에 있어 배치 정규화 방법이 추가된 CNN을 이용하여 영상 분류의 정확도를 높이는 방법을 제시하였다. 본 논문에서는 영상 분류를 더 정확하게 수행하기 위해 기존의 뉴럴 네트워크에 배치 정규화 계층 (layer)를 추가하는 방법을 제안한다. 배치 정규화는 각 계층에 존재하는 편향을 줄이기 위해 고안된 방법으로, 각 배치의 평균과 분산을 계산하여 이동시키는 방법이다. 본 논문에서 제시된 방법의 우수성을 입증하기 위하여 SHREC13, MNIST, SVHN, CIFAR-10, CIFAR-100의 5개 영상 데이터 집합을 이용하여 영상분류 실험을 하여 정확도와 mAP를 측정한다. 실험 결과 일반적인 CNN 보다 배치 정규화가 추가된 CNN이 영상 분류 시 보다 높은 분류 정확도와 mAP를 보임을 확인 할 수 있었다.

• Nuclear Engineering and Technology
• /
• 제53권2호
• /
• pp.522-531
• /
• 2021

We describe a new deep learning model - Convolutional eXtreme Gradient Boosting (ConvXGB) for classification problems based on convolutional neural nets and Chen et al.'s XGBoost. As well as image data, ConvXGB also supports the general classification problems, with a data preprocessing module. ConvXGB consists of several stacked convolutional layers to learn the features of the input and is able to learn features automatically, followed by XGBoost in the last layer for predicting the class labels. The ConvXGB model is simplified by reducing the number of parameters under appropriate conditions, since it is not necessary re-adjust the weight values in a back propagation cycle. Experiments on several data sets from UCL Repository, including images and general data sets, showed that our model handled the classification problems, for all the tested data sets, slightly better than CNN and XGBoost alone and was sometimes significantly better.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.383-392
• /
• 2023

Post-earthquake building condition assessment is crucial for subsequent rescue and remediation and can be automated by emerging computer vision and deep learning technologies. This study is based on an endeavour for the 2nd International Competition of Structural Health Monitoring (IC-SHM 2021). The task package includes five image segmentation objectives - defects (crack/spall/rebar exposure), structural component, and damage state. The structural component and damage state tasks are identified as the priority that can form actionable decisions. A multi-task Convolutional Neural Network (CNN) is proposed to conduct the two major tasks simultaneously. The rest 3 sub-tasks (spall/crack/rebar exposure) were incorporated as auxiliary tasks. By synchronously learning defect information (spall/crack/rebar exposure), the multi-task CNN model outperforms the counterpart single-task models in recognizing structural components and estimating damage states. Particularly, the pixel-level damage state estimation witnesses a mIoU (mean intersection over union) improvement from 0.5855 to 0.6374. For the defect detection tasks, rebar exposure is omitted due to the extremely biased sample distribution. The segmentations of crack and spall are automated by single-task U-Net but with extra efforts to resample the provided data. The segmentation of small objects (spall and crack) benefits from the resampling method, with a substantial IoU increment of nearly 10%.

• 한국멀티미디어학회논문지
• /
• 제21권10호
• /
• pp.1142-1149
• /
• 2018

Recently, computer vision application is increasing by using CNN which is one of the deep learning algorithms. However, CNN does not provide perfect classification performance due to gradient vanishing problem. Most of CNN algorithms use an activation function called ReLU to mitigate the gradient vanishing problem. In this study, four activation functions that can replace ReLU were applied to four different structural networks. Experimental results show that ReLU has the lowest performance in accuracy, loss rate, and speed of initial learning convergence from 20 experiments. It is concluded that the optimal activation function varied from network to network but the four activation functions were higher than ReLU.

• 반도체디스플레이기술학회지
• /
• 제20권4호
• /
• pp.151-156
• /
• 2021

Low-end IoT devices do not have enough computation and memory resources for DNN learning and inference. Integer quantization of real-type neural network models can reduce model size, hardware computational burden, and power consumption. This paper describes the design and implementation of a web-based quantization platform for CNN deep learning accelerator chips. In the web service platform, we implemented visualization of the model through a convenient UI, analysis of each step of inference, and detailed editing of the model. Additionally, a data augmentation function and a management function of files that store models and inference intermediate results are provided. The implemented functions were verified using three YOLO models.

• International Journal of Computer Science & Network Security
• /
• 제24권5호
• /
• pp.129-134
• /
• 2024

Object Detection is an emerging technology in the field of Computer Vision and Image Processing that deals with detecting objects of a particular class in digital images. It has considered being one of the complicated and challenging tasks in computer vision. Earlier several machine learning-based approaches like SIFT (Scale-invariant feature transform) and HOG (Histogram of oriented gradients) are widely used to classify objects in an image. These approaches use the Support vector machine for classification. The biggest challenges with these approaches are that they are computationally intensive for use in real-time applications, and these methods do not work well with massive datasets. To overcome these challenges, we implemented a Deep Learning based approach Convolutional Neural Network (CNN) in this paper. The Proposed approach provides accurate results in detecting objects in an image by the area of object highlighted in a Bounding Box along with its accuracy.

• 전기전자학회논문지
• /
• 제22권4호
• /
• pp.1242-1245
• /
• 2018

본 논문에서, Gaussian noise를 제거할 때 발생하는 over blurring 현상을 감소시키는 network를 구현하였다. 기존 filtering 방식은 원 영상을 blurring하여 noise를 제거함으로써, edge나 corner 같은 high frequency 성분도 함께 지워지는 것을 확인할 수 있다. CNN (Convolutional Neural Network)기반 denoiser의 경우도 사소한 edge, keypoint를 noise로 인식하여 이러한 정보를 잃게 된다. 우리는 CNN을 기반으로 denoising된 high frequency 성분만을 획득하여 기존 denoiser에 추가함으로써 denoising 성능을 유지하면서 over blurring을 완화하는 network 제안한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권7호
• /
• pp.2752-2768
• /
• 2020

Phishing websites can have devastating effects on governmental, financial, and social services, as well as on individual privacy. Currently, many phishing detection solutions are evaluated using small datasets and, thus, are prone to sampling issues, such as representing legitimate websites by only high-ranking websites, which could make their evaluation less relevant in practice. Phishing detection solutions which depend only on the URL are attractive, as they can be used in limited systems, such as with firewalls. In this paper, we present a URL-only phishing detection solution based on a convolutional neural network (CNN) model. The proposed CNN takes the URL as the input, rather than using predetermined features such as URL length. For training and evaluation, we have collected over two million URLs in a massive URL phishing detection (MUPD) dataset. We split MUPD into training, validation and testing datasets. The proposed CNN achieves approximately 96% accuracy on the testing dataset; this accuracy is achieved with URL schemes (such as HTTP and HTTPS) removed from the URL. Our proposed solution achieved better accuracy compared to an existing state-of-the-art URL-only model on a published dataset. Finally, the results of our experiment suggest keeping the CNN up-to-date for better results in practice.

• Current Optics and Photonics
• /
• 제6권1호
• /
• pp.92-103
• /
• 2022

The development of midinfrared (mid-IR) quantum cascade lasers (QCLs) has enabled rapid high-contrast measurement of the mid-IR spectra of biological tissues. Several studies have compared the differences between the mid-IR spectra of colon cancer and noncancerous colon tissues. Most mid-IR spectrum classification studies have been proposed as machine-learning-based algorithms, but this results in deviations depending on the initial data and threshold values. We aim to develop a process for classifying colon cancer and noncancerous colon tissues through a deep-learning-based convolutional-neural-network (CNN) model. First, we image the midinfrared spectrum for the CNN model, an image-based deep-learning (DL) algorithm. Then, it is trained with the CNN algorithm and the classification ratio is evaluated using the test data. When the tissue microarray (TMA) and routine pathological slide are tested, the ML-based support-vector-machine (SVM) model produces biased results, whereas we confirm that the CNN model classifies colon cancer and noncancerous colon tissues. These results demonstrate that the CNN model using midinfrared-spectrum images is effective at classifying colon cancer tissue and noncancerous colon tissue, and not only submillimeter-sized TMA but also routine colon cancer tissue samples a few tens of millimeters in size.

• 한국전자통신학회논문지
• /
• 제18권6호
• /
• pp.1259-1268
• /
• 2023

본 연구는 딥러닝의 합성곱과 순환신경망 네트워크를 기반으로 시각센서를 이용해 속도(Throttle)와 조향(Steering) 제어 기술을 제안한다. 학습 트랙을 시계, 반시계 방향으로 주행하며 카메라 영상 이미지와 조종 값 데이터를 수집하고 효율적인 학습을 위해 데이터 샘플링, 전처리 과정을 거쳐 Throttle과 Steering을 예측하는 모델을 생성한다. 이후 학습에 사용되지 않은 다른 환경의 테스트 트랙을 통해 검증을 진행하여 최적의 모델을 찾고 이를 CNN(Convolutional Neural Network)과 비교하였다. 그 결과 제안하는 딥러닝 모델의 성능이 뛰어남을 확인했다.