• International Journal of Internet, Broadcasting and Communication
• /
• v.11 no.4
• /
• pp.37-42
• /
• 2019

In this paper, we explore the details of three classic data augmentation methods and two generative model based oversampling methods. The three classic data augmentation methods are random sampling (RANDOM), Synthetic Minority Over-sampling Technique (SMOTE), and Adaptive Synthetic Sampling (ADASYN). The two generative model based oversampling methods are Conditional Generative Adversarial Network (CGAN) and Wasserstein Generative Adversarial Network (WGAN). In imbalanced data, the whole instances are divided into majority class and minority class, where majority class occupies most of the instances in the training set and minority class only includes a few instances. Generative models have their own advantages when they are used to generate more plausible samples referring to the distribution of the minority class. We also adopt CGAN to compare the data augmentation performance with other methods. The experimental results show that WGAN-based oversampling technique is more stable than other approaches (RANDOM, SMOTE, ADASYN and CGAN) even with the very limited training datasets. However, when the imbalanced ratio is too small, generative model based approaches cannot achieve satisfying performance than the conventional data augmentation techniques. These results suggest us one of future research directions.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.479-484
• /
• 1998

입력된 영상으로부터 추론된 정보 표를 지식베이스에 저장하여 결정해를 구하는데는 많은 탐색시간이 소비된다. 본 논문에서는 탐색 시간을 감소시키기 위해서 러프집합의 식별(classification)과 근사(approximation) 개념을 이용하여 추론된 정보를 동치 클래스(equivalence class)로 분류하여 간략화한다. 감소된 규칙, 즉 Core와 Reduct 리스트를 구하여 결정해를 검색하는데 탐색 시간을 감소시키는데 있다.

• Imaging Science in Dentistry
• /
• v.52 no.1
• /
• pp.75-82
• /
• 2022

Purpose: This cross-sectional study evaluated and categorized the tomographic sagittal root position (SRP) of the maxillary anterior teeth in a Brazilian population. Materials and Methods: Cone-beam computed tomographic scans of 420 maxillary anterior teeth of 70 patients (35 men and 35 women, mean age 25.2±5.9 years) were evaluated. The SRP was classified as class I, II, III, or IV. In class I, the root is positioned against the buccal cortical plate; in class II, the root is centered in the middle of the alveolar housing; in class III, the root is positioned against the palatal cortical plate; and in class IV, at least two-thirds of the root engage both the buccal and palatal cortical plates. Results: In total, 274 teeth (65.2%) were class I, 39 (9.3%) were class II, 3 (0.7%) were class III, and 104 (24.8%) were class IV. The frequency distribution over the teeth groups was different from the overall analysis. Important differences were found in the frequencies of classes I, II, and IV compared to other populations. Sex was not associated with the SRP classes (P=0.307). Age distribution was significantly different over the classes (P=0.004). Conclusion: The findings of this study on the distribution of SRP classes among the Brazilian population compared to other populations demonstrate that the SRP should be analyzed on a case-by-case basis for an accurate treatment plan in the maxillary anterior area.

• ETRI Journal
• /
• v.42 no.1
• /
• pp.67-77
• /
• 2020

Image captioning has received significant interest in recent years, and notable results have been achieved. Most previous approaches have focused on generating visual descriptions from images, whereas a few approaches have exploited visual descriptions for image classification. This study demonstrates that a good performance can be achieved for both description generation and image classification through an end-to-end joint learning approach with a loss function, which encourages each task to reach a consensus. When given images and visual descriptions, the proposed model learns a multimodal intermediate embedding, which can represent both the textual and visual characteristics of an object. The performance can be improved for both tasks by sharing the multimodal embedding. Through a novel loss function based on class activation mapping, which localizes the discriminative image region of a model, we achieve a higher score when the captioning and classification model reaches a consensus on the key parts of the object. Using the proposed model, we established a substantially improved performance for each task on the UCSD Birds and Oxford Flowers datasets.

• Korean Journal of Remote Sensing
• /
• v.24 no.1
• /
• pp.57-64
• /
• 2008

In the previous study, the Haar wavelet was used as the sole basis function for the 3D discrete wavelet transform because the number of bands is too small to decompose a remotely sensed image in band direction with other basis functions. However, it is possible to use other basis functions for wavelet decomposition in horizontal and vertical directions because wavelet decomposition is independently performed in each direction. This study aims to classify a high spatial resolution image with the six types of basis function including the Haar function and to compare those results. The other wavelets are more helpful to classify high resolution imagery than the Haar wavelet. In overall accuracy, the Coif4 wavelet has the best result. The improvement of classification accuracy is different depending on the type of class and the type of wavelet. Using the basis functions with long length could be effective for improving accuracy in classification, especially for the classes of small area. This study is expected to be used as fundamental information for selecting optimal basis function according to the data properties in the 3D DWT based image classification.

• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• pp.690-706
• /
• 2021

Classification is a substantial operation in data mining, and each element is distributed taking into account its feature values in the corresponding class. Metaheuristics have been widely used in attempts to solve satellite image classification problems. This article proposes a hybrid approach, the flower pigeons-inspired optimization algorithm (FPIO), and the local search method of the flower pollination algorithm is integrated into the pigeon-inspired algorithm. The efficiency and power of the proposed FPIO approach are displayed with a series of images, supported by computational results that demonstrate the cogency of the proposed classification method on satellite imagery. For this work, the Davies-Bouldin Index is used as an objective function. FPIO is applied to different types of images (synthetic, Alsat-2B, and Sentinel-2). Moreover, a comparative experiment between FPIO and the genetic algorithm genetic algorithm is conducted. Experimental results showed that GA outperformed FPIO in matters of time computing. However, FPIO provided better quality results with less confusion. The overall experimental results demonstrate that the proposed approach is an efficient method for satellite imagery classification.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_4
• /
• pp.1911-1923
• /
• 2022

The availability of high-resolution satellite images provides precise information without direct observation of the research target. Korea Multi-Purpose Satellite (KOMPSAT), also known as the Arirang satellite, has been developed and utilized for earth observation. The machine learning model was continuously proven as a good classifier in classifying remotely sensed images. This study aimed to compare the performance of the support vector machine (SVM) model in classifying the land cover of the Delaware River port area on high and medium-resolution images. Three optical images, which are KOMPSAT-2, KOMPSAT-3A, and Sentinel-2B, were classified into six land cover classes, including water, road, vegetation, building, vacant, and shadow. The KOMPSAT images are provided by Korea Aerospace Research Institute (KARI), and the Sentinel-2B image was provided by the European Space Agency (ESA). The training samples were manually digitized for each land cover class and considered the reference image. The predicted images were compared to the actual data to obtain the accuracy assessment using a confusion matrix analysis. In addition, the time-consuming training and classifying were recorded to evaluate the model performance. The results showed that the KOMPSAT-3A image has the highest overall accuracy and followed by KOMPSAT-2 and Sentinel-2B results. On the contrary, the model took a long time to classify the higher-resolution image compared to the lower resolution. For that reason, we can conclude that the SVM model performed better in the higher resolution image with the consequence of the longer time-consuming training and classifying data. Thus, this finding might provide consideration for related researchers when selecting satellite imagery for effective and accurate image classification.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.3
• /
• pp.1243-1263
• /
• 2018

The two-stage linear discrimination analysis (TSLDA) is a feature extraction technique to solve the small size sample problem in the field of image recognition. The TSLDA has retained all subspace information of the between-class scatter and within-class scatter. However, the feature information in the four subspaces may not be entirely beneficial for classification, and the regularization procedure for eliminating singular metrics in TSLDA has higher time complexity. In order to address these drawbacks, this paper proposes an improved two-stage linear discriminant analysis (Improved TSLDA). The Improved TSLDA proposes a selection and compression method to extract superior feature information from the four subspaces to constitute optimal projection space, where it defines a single Fisher criterion to measure the importance of single feature vector. Meanwhile, Improved TSLDA also applies an approximation matrix method to eliminate the singular matrices and reduce its time complexity. This paper presents comparative experiments on five face databases and one handwritten digit database to validate the effectiveness of the Improved TSLDA.

• Journal of Science Education
• /
• v.44 no.1
• /
• pp.61-73
• /
• 2020

Teachers are the most important factor contributing to determining the quality of education. Therefore, the quality of teachers should be improved to enhance the quality of education. Teacher's competencies are defined as the skills required for teaching profession, that is, the ability to perform not only in teaching activities, but also in guidance and class management. The purpose of this study is to analyze the competencies of science teachers that elementary, middle and high school students want. To this end, 332 elementary, middle and high school students were asked to describe their preferred science teacher's competencies and avoiding science teacher's competencies as an open questionnaire. The resulting concepts were analyzed by semantic network analysis (SNA). The results of this study are as follows: 1) The competencies of science teachers that students prefer varied. This suggests that most students think positively about science teachers. In addition, it is possible to show students the positive or preferred competencies of teachers in various ways. 2) The students wanted teachers to explain the theories and concepts related to scientific phenomena through experiments. They also preferred hands-on activities and experience in science class. 3) The students put emphasis on the class-related contents in the competencies of science teachers. Accordingly, the image of science teachers and science itself should be enhanced through the improvement of science teaching methods and positive attitudes toward students. It is expected that further research on the image according to specific teaching methods of science teachers will be conducted based on the findings of this study.

• Culinary science and hospitality research
• /
• v.10 no.4
• /
• pp.50-66
• /
• 2004

We defined life style as something that every members of society have in common. These social and cultural environments build up not only society group or every individual's expectation but also its own life style. In that way, these social and cultural environments leads to particular consumer behavior pattern in this food-service industry. So we regard next generation's trend which consists of rational consumers as important indicator when we make future's plan in foodservice industry. We consider smart map which needs rational and continuous consume pattern as the construction of next generation's main consumer class. Therefore, this study tried to develop of color and image marketing strategies to attract LOHAS and nomadic consumer.