• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.9
• /
• pp.1049-1059
• /
• 2019

We propose a learning-based image segmentation algorithm. Starting from super-pixels, our method learns the probability of merging two regions based on the ground truth made by humans. The learned information is used in determining whether the two regions should be merged or not in a segmentation stage. Unlike exiting learning-based algorithms, we use both local and object information. The local information represents features computed from super-pixels and the object information represent high level information available only in the learning process. The object information is considered as privileged information, and we can use a framework that utilize the privileged information such as SVM+. In experiments on the Berkeley Segmentation Dataset and Benchmark (BSDS 500) and PASCAL Visual Object Classes Challenge (VOC 2012) data set, out model exhibited the best performance with a relatively small training data set and also showed competitive results with a sufficiently large training data set.

• The Journal of Society for e-Business Studies
• /
• v.25 no.3
• /
• pp.1-13
• /
• 2020

Machine learning refers to a model generation technique that can solve specific problems from the generalization process for given data. In order to generate a high performance model, high quality training data and learning algorithms for generalization process should be prepared. As one way of improving the performance of model to be learned, the Ensemble technique generates multiple models rather than a single model, which includes bagging, boosting, and stacking learning techniques. This paper proposes a new Ensemble technique with multiple stacking that outperforms the conventional stacking technique. The learning structure of multiple stacking ensemble technique is similar to the structure of deep learning, in which each layer is composed of a combination of stacking models, and the number of layers get increased so as to minimize the misclassification rate of each layer. Through experiments using four types of datasets, we have showed that the proposed method outperforms the exiting ones.

• Journal of the Korea Convergence Society
• /
• v.6 no.4
• /
• pp.113-119
• /
• 2015

Currently, we are presenting an alternative educational environment for the normal student of education rules failure of the only that has been the school system student. The study for students with disabilities, it is designed especially to be able to use difficult disabilities the use of hand. Development objectives of the learning video e-learning system of persons with disabilities, is that to be able to capable of self-directed learning of disabled students. Configuration of e-running system, Web-based multimedia system, utilizing the system that will change the video conferencing system and voice to a character hearing impaired students through the chat system is 1:1 by communication, and teachers it is possible to perform two-way communication. A learning disability e-learning system developed in this paper between teachers and students with disabilities 1:1 training is conducted using a two-way communication algorithms.

• The Journal of Information Systems
• /
• v.29 no.3
• /
• pp.35-52
• /
• 2020

Purpose Local governments in each region actively hold local festivals for the purpose of promoting the region and revitalizing the local economy. Existing studies related to local festivals have been actively conducted in tourism and related academic fields. Empirical studies to understand the effects of latent variables on local festivals and studies to analyze the regional economic impacts of festivals occupy a large proportion. Despite of practical need, since few researches have been conducted to predict the number of visitors, one of the criteria for evaluating the performance of local festivals, this study developed a model for predicting the number of visitors through various observed variables using a machine learning algorithm and derived its implications. Design/methodology/approach For a total of 593 festivals held in 2018, 6 variables related to the region considering population size, administrative division, and accessibility, and 15 variables related to the festival such as the degree of publicity and word of mouth, invitation singer, weather and budget were set for the training data in machine learning algorithm. Since the number of visitors is a continuous numerical data, random forest, Adaboost, and linear regression that can perform regression analysis among the machine learning algorithms were used. Findings This study confirmed that a prediction of the number of visitors to local festivals is possible using a machine learning algorithm, and the possibility of using machine learning in research in the tourism and related academic fields, including the study of local festivals, was captured. From a practical point of view, the model developed in this study is used to predict the number of visitors to the festival to be held in the future, so that the festival can be evaluated in advance and the demand for related facilities, etc. can be utilized. In addition, the RReliefF rank result can be used. Considering this, it will be possible to improve the existing local festivals or refer to the planning of a new festival.

• Journal of Digital Contents Society
• /
• v.15 no.6
• /
• pp.737-743
• /
• 2014

For single image super resolution (SR), interpolation based and example based algorithms are extensively used. The interpolation algorithms have the strength of theoretical simplicity. However, those algorithms are tending to produce high resolution images with jagged edges, because they are not able to use more priori information. Example based algorithms have been studied in the past few years. For example based SR, the nearest neighbor based algorithms are extensively considered. Among them, neighbor embedding (NE) has been inspired by manifold learning method, particularly locally linear embedding. However, the sizes of local training sets are always too small. So, NE algorithm is weak in the performance of the visuality and quantitative measure by the poor generalization of nearest neighbor estimation. An improved NE algorithm with Support Vector Regression (SVR) was proposed to solve this problem. Given a low resolution image, the pixel values in its high resolution version are estimated by the improved NE. Comparing with bicubic and NE, the improvements of 1.25 dB and 2.33 dB are achieved in PSNR. Experimental results show that proposed method is quantitatively and visually more effective than prior works using bicubic interpolation and NE.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.28 no.4
• /
• pp.55-61
• /
• 2024

Recently, the number of aging concrete structures is steadily increasing. This is because many of these structures are reaching their expected lifespan. Such structures require accurate inspections and persistent maintenance. Otherwise, their original functions and performance may degrade, potentially leading to safety accidents. Therefore, research on objective inspection technologies using deep learning and computer vision is actively being conducted. High-resolution images can accurately observe not only micro cracks but also spalling and exposed rebar, and deep learning enables automated detection. High detection performance in deep learning is only guaranteed with diverse and numerous training datasets. However, surface damage to concrete is not commonly captured in images, resulting in a lack of training data. To overcome this limitation, this study proposed a method for generating concrete surface damage images, including cracks, spalling, and exposed rebar, using stable diffusion. This method synthesizes new damage images by paired text and image data. For this purpose, a training dataset of 678 images was secured, and fine-tuning was performed through low-rank adaptation. The quality of the generated images was compared according to three base models of stable diffusion. As a result, a method to synthesize the most diverse and high-quality concrete damage images was developed. This research is expected to address the issue of data scarcity and contribute to improving the accuracy of deep learning-based damage detection algorithms in the future.

• Korean Journal of Remote Sensing
• /
• v.37 no.5_1
• /
• pp.871-884
• /
• 2021

The purpose of this study was to determine ways to increase efficiency in constructing and verifying artificial intelligence learning data on land cover using aerial and satellite images, and in applying the data to AI learning algorithms. To this end, multi-resolution datasets of 0.51 m and 10 m each for 8 categories of land cover were constructed using high-resolution aerial images and satellite images obtained from Sentinel-2 satellites. Furthermore, fine data (a total of 17,000 pieces) and coarse data (a total of 33,000 pieces) were simultaneously constructed to achieve the following two goals: precise detection of land cover changes and the establishment of large-scale learning datasets. To secure the accuracy of the learning data, the verification was performed in three steps, which included data refining, annotation, and sampling. The learning data that wasfinally verified was applied to the semantic segmentation algorithms U-Net and DeeplabV3+, and the results were analyzed. Based on the analysis, the average accuracy for land cover based on aerial imagery was 77.8% for U-Net and 76.3% for Deeplab V3+, while for land cover based on satellite imagery it was 91.4% for U-Net and 85.8% for Deeplab V3+. The artificial intelligence learning datasets on land cover constructed using high-resolution aerial and satellite images in this study can be used as reference data to help classify land cover and identify relevant changes. Therefore, it is expected that this study's findings can be used in the future in various fields of artificial intelligence studying land cover in constructing an artificial intelligence learning dataset on land cover of the whole of Korea.

• Smart Media Journal
• /
• v.11 no.2
• /
• pp.31-38
• /
• 2022

Promptly predicting changes in the salinity in rivers is an important task to predict the damage to agriculture and ecosystems caused by salinity infiltration and to establish disaster prevention measures. Because machine learning(ML) methods show much less computation cost than physics-based hydraulic models, they can predict the river salinity in a relatively short time. Due to shorter training time, ML methods have been studied as a complementary technique to physics-based hydraulic model. Many studies on salinity prediction based on machine learning have been studied actively around the world, but there are few studies in South Korea. With a massive number of datasets available publicly, we evaluated the performance of various kinds of machine learning techniques that predict the salinity of the Nakdong River Estuary Basin. As a result, LightGBM algorithm shows average 0.37 in RMSE as prediction performance and 2-20 times faster learning speed than other algorithms. This indicates that machine learning techniques can be applied to predict the salinity of rivers in Korea.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.25 no.2
• /
• pp.80-98
• /
• 2023

This study explores the effectiveness of various data preprocessing algorithms for improving subseasonal to seasonal (S2S) climate predictions from six climate forecast models and their Multi-Model Ensemble (MME) using a deep learning-based postprocessing model. A pipeline of data transformation algorithms was constructed to convert raw S2S prediction data into the training data processed with several statistical distribution. A dimensionality reduction algorithm for selecting features through rankings of correlation coefficients between the observed and the input data. The training model in the study was designed with TimeDistributed wrapper applied to all convolutional layers of U-Net: The TimeDistributed wrapper allows a U-Net convolutional layer to be directly applied to 5-dimensional time series data while maintaining the time axis of data, but every input should be at least 3D in U-Net. We found that Robust and Standard transformation algorithms are most suitable for improving S2S predictions. The dimensionality reduction based on feature selections did not significantly improve predictions of daily precipitation for six climate models and even worsened predictions of daily maximum and minimum temperatures. While deep learning-based postprocessing was also improved MME S2S precipitation predictions, it did not have a significant effect on temperature predictions, particularly for the lead time of weeks 1 and 2. Further research is needed to develop an optimal deep learning model for improving S2S temperature predictions by testing various models and parameters.

• Journal of Digital Contents Society
• /
• v.18 no.6
• /
• pp.1165-1173
• /
• 2017

Deep learning method, which is one of the most famous machine learning algorithms, has proven its applicability in various applications and is widely used in digital signal processing. However, it is difficult to apply deep learning technology to IoT devices with limited CPU performance and memory capacity, because a large number of training samples requires a lot of memory and computation time. In particular, if the Arduino with a very small memory capacity of 2K to 8K, is used, there are many limitations in implementing the algorithm. In this paper, we propose a method to optimize the ELM algorithm, which is proved to be accurate and efficient in various fields, on Arduino board. Experiments have shown that multi-class learning is possible up to 15-dimensional data on Arduino UNO with memory capacity of 2KB and possible up to 42-dimensional data on Arduino MEGA with memory capacity of 8KB. To evaluate the experiment, we proved the effectiveness of the proposed algorithm using the data sets generated using gaussian mixture modeling and the public UCI data sets.