• 반도체디스플레이기술학회지
• /
• 제19권2호
• /
• pp.60-67
• /
• 2020

Recently, artificial intelligence related technologies including machine learning are being applied to various fields, and the demand is also increasing. In particular, with the development of AR, VR, and MR technologies related to image processing, the utilization of computer vision based on deep learning has increased. The algorithms for object recognition and detection based on deep learning required for image processing are diversified and advanced. Accordingly, problems that were difficult to solve with the existing methodology were solved more simply and easily by using deep learning. This paper introduces various deep learning-based object recognition and extraction algorithms used to detect and recognize various objects in an image and analyzes the technologies that attract attention.

• 한국군사과학기술학회지
• /
• 제22권1호
• /
• pp.49-59
• /
• 2019

Generative adversarial networks(GANs) have received great attention in the machine learning field for their capacity to model high-dimensional and complex data distribution implicitly and generate new data samples from the model distribution. This paper investigates the model training methodology, architecture, and various applications of generative adversarial networks. Experimental evaluation is also conducted for generating synthetic image dataset for defense using two types of GANs. The first one is for military image generation utilizing the deep convolutional generative adversarial networks(DCGAN). The other is for visible-to-infrared image translation utilizing the cycle-consistent generative adversarial networks(CycleGAN). Each model can yield a great diversity of high-fidelity synthetic images compared to training ones. This result opens up the possibility of using inexpensive synthetic images for training neural networks while avoiding the enormous expense of collecting large amounts of hand-annotated real dataset.

• International Journal of Computer Science & Network Security
• /
• 제21권6호
• /
• pp.137-142
• /
• 2021

Machine-learning systems have proven their worth in various industries, including healthcare and banking, by assisting in the extraction of valuable inferences. Information in these crucial sectors is traditionally stored in databases distributed across multiple environments, making accessing and extracting data from them a tough job. To this issue, we must add that these data sources contain sensitive information, implying that the data cannot be shared outside of the head. Using cryptographic techniques, Privacy-Preserving Machine Learning (PPML) helps solve this challenge, enabling information discovery while maintaining data privacy. In this paper, we talk about how to keep your data mining private. Because Data mining has a wide variety of uses, including business intelligence, medical diagnostic systems, image processing, web search, and scientific discoveries, and we discuss privacy-preserving in deep learning because deep learning (DL) exhibits exceptional exactitude in picture detection, Speech recognition, and natural language processing recognition as when compared to other fields of machine learning so that it detects the existence of any error that may occur to the data or access to systems and add data by unauthorized persons.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권4호
• /
• pp.1738-1764
• /
• 2019

With the advent of big data, deep learning technology has become an important research direction in the field of machine learning, which has been widely applied in the image processing, natural language processing, speech recognition and online advertising and so on. This paper introduces deep learning techniques from various aspects, including common models of deep learning and their optimization methods, commonly used open source frameworks, existing problems and future research directions. Firstly, we introduce the applications of deep learning; Secondly, we introduce several common models of deep learning and optimization methods; Thirdly, we describe several common frameworks and platforms of deep learning; Finally, we introduce the latest acceleration technology of deep learning and highlight the future work of deep learning.

• 스마트미디어저널
• /
• 제11권7호
• /
• pp.61-69
• /
• 2022

학습된 머신러닝은 시간 경과에 따른 학습 모델과 학습 데이터 측면의 표류 현상이 발생과 동시에 머신러닝의 성능이 퇴화하게 된다. 이를 해결하기 위한 방안으로 머신러닝의 재학습 시기를 결정하기 위한 ML 표류의 개념과 평가 방법을 제안하고자 한다. 딸기와 선명도에 따른 XAI 테스트 및 사과 이미지의 XAI 테스트를 진행하였다. 딸기의 경우 선명도 값에 따른 ML 모델의 XAI 분석의 변화는 미미하였으며 사과 이미지의 XAI의 경우 사과는 정상적으로 객체 분류 및 히트맵 영역을 표시하였으나 사과꽃 및 꽃봉오리의 경우 그 결과가 딸기나 사과에 비해 미미하였다. 이는 사과꽃 및 꽃봉오리의 학습 이미지 수가 부족하기에 발생한 것으로 예상되며 추후 더 많은 사과꽃 및 꽃봉오리 이미지를 학습하여 테스트할 계획이다.

• Journal of Information Processing Systems
• /
• 제13권2호
• /
• pp.204-214
• /
• 2017

Artificial intelligence, especially deep learning technology, is penetrating the majority of research areas, including the field of bioinformatics. However, deep learning has some limitations, such as the complexity of parameter tuning, architecture design, and so forth. In this study, we analyze these issues and challenges in regards to its applications in bioinformatics, particularly genomic analysis and medical image analytics, and give the corresponding approaches and solutions. Although these solutions are mostly rule of thumb, they can effectively handle the issues connected to training learning machines. As such, we explore the tendency of deep learning technology by examining several directions, such as automation, scalability, individuality, mobility, integration, and intelligence warehousing.

• 한국멀티미디어학회논문지
• /
• 제25권1호
• /
• pp.52-60
• /
• 2022

Few-shot learning is a sub-area of machine learning problems, which aims to classify target images that only contain a few labeled samples for training. As a representative few-shot learning method, the Prototypical network has been received much attention due to its simplicity and promising results. However, the Prototypical network uses the sample mean of samples from the same class as the prototypes of that class, which easily results in learning uncharacteristic features in the low-data scenery. In this study, we propose to use local descriptors (i.e., patches along the channel within feature maps) from the same class to explicitly obtain more representative prototypes for Prototypical Network so that significant intra-class feature information can be maintained and thus improving the classification performance on few-shot learning tasks. Experimental results on various benchmark datasets including mini-ImageNet, CUB-200-2011, and tiered-ImageNet show that the proposed method can learn more discriminative intra-class features by the local descriptors and obtain more generic prototype representations under the few-shot setting.

• 한국항행학회논문지
• /
• 제24권5호
• /
• pp.430-437
• /
• 2020

제안 모델은 얼굴 검출과 랜드마크 및 얼굴 인식 알고리즘을 이용하여 인공신경망으로 학습을 통해 얼굴 예측률과 인식률을 향상하는 모델을 구현하였다. 제안 모델은 특정 인물의 얼굴 영상에서 랜드마킹을 한 후, 기존에 학습된 Caffe 모델을 이용하여 얼굴검출과 임베딩 벡터 128D를 추출하였다. 학습은 기계학습 알고리즘인 SVM (support vector machine)과 DNN (deep neural network)을 구축하여 학습하였다. 얼굴인식은 학습된 모델을 이용하여 학습된 인물 중 다른 얼굴 영상으로 테스트하였다. 실험 결과, SVM 보다는 DNN으로 학습한 결과가 우수한 예측률과 인식률을 보였다. DNN의 중간층을 증가하게 되면 예측률은 높아지나 인식률이 감소하는 현상이 발생하였다. 이것은 인식하고자 하는 대상이 적음으로써 발생하는 과적합으로 판단된다. 제안 모델은 명확한 얼굴 영상을 추가하여 학습한 결과, 높은 예측률과 인식률의 결과를 얻을 수 있음을 확인할 수 있었다. 본 연구는 좀 더 많은 얼굴 영상 데이터를 이용함으로써 보다 효과적인 딥러닝 구축을 통해 보다 향상된 인식률과 예측률을 얻을 수 있을 것이다.

• International Journal of Internet, Broadcasting and Communication
• /
• 제14권1호
• /
• pp.136-141
• /
• 2022

A very complex task in deep learning such as image classification must be solved with the help of neural networks and activation functions. The backpropagation algorithm advances backward from the output layer towards the input layer, the gradients often get smaller and smaller and approach zero which eventually leaves the weights of the initial or lower layers nearly unchanged, as a result, the gradient descent never converges to the optimum. We propose a two-factor non-saturating activation functions known as Bea-Mish for machine learning applications in deep neural networks. Our method uses two factors, beta (𝛽) and alpha (𝛼), to normalize the area below the boundary in the Mish activation function and we regard these elements as Bea. Bea-Mish provide a clear understanding of the behaviors and conditions governing this regularization term can lead to a more principled approach for constructing better performing activation functions. We evaluate Bea-Mish results against Mish and Swish activation functions in various models and data sets. Empirical results show that our approach (Bea-Mish) outperforms native Mish using SqueezeNet backbone with an average precision (AP50val) of 2.51% in CIFAR-10 and top-1accuracy in ResNet-50 on ImageNet-1k. shows an improvement of 1.20%.

• 한국운동역학회지
• /
• 제17권1호
• /
• pp.81-90
• /
• 2007

The effort identifying positioning information of the moving object in real time has been a issue not only in sport biomechanics but also other academic areas. In order to solve this issue, this study tried to track the movement of a pitched ball that might provide an easier prediction because of a clear focus and simple movement of the object. Machine learning has been leading the research of extracting information from continuous images such as object tracking. Though the rule-based methods in artificial intelligence prevailed for decades, it has evolved into the methods of statistical approach that finds the maximum a posterior location in the image. The development of machine learning, accompanied by the development of recording technology and computational power of computer, made it possible to extract the trajectory of pitched baseball from recorded images. We present a method of baseball tracking, based on object tracking methods in machine learning. We introduce three state-of-the-art researches regarding the object tracking and show how we can combine these researches to yield a novel engine that finds trajectory from continuous pitching images. The first research is about mean shift method which finds the mode of a supposed continuous distribution from a set of data. The second research is about the research that explains how we can find the mode and object region effectively when we are given the previous image's location of object and the region. The third is about the research of representing data into features that we can deal with. From those features, we can establish a distribution to generate a set of data for mean shift. In this paper, we combine three works to track baseball's location in the continuous image frames. From the information of locations from two sets of images, we can reconstruct the real 3-D trajectory of pitched ball. We show how this works in real pitching images.