• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2008.08a
• /
• pp.142-145
• /
• 2008

This paper proposes a computationally efficient algorithm for color image segmentation using self-organizing map(SOM) and grassfire algorithm. We reduce a computation time by decreasing the number of input neuron and input data which is used for learning at SOM. First converting input image to CIE $L^*u^*v^*$ color space and run the learning stage with the SOM-input neuron size is three and output neuron structure is 4by4 or 5by5. After learning, compute output value correspondent with input pixel and merge adjacent pixels which have same output value into segment using grassfire algorithm. The experimental results with various images show that proposed method lead to a good segmentation results than others.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.1
• /
• pp.65-70
• /
• 2023

In this paper, we used a deep learning model to detect and segment flame and smoke in real time from fires. To this end, well known U-NET was used to separate and divide the flame and smoke of the fire using multi-class. As a result of learning using the proposed technique, the values of loss error and accuracy are very good at 0.0486 and 0.97996, respectively. The IOU value used in object detection is also very good at 0.849. As a result of predicting fire images that were not used for learning using the learned model, the flame and smoke of fire are well detected and segmented, and smoke color were well distinguished. Proposed method can be used to build fire prediction and detection system.

• Journal of Korea Game Society
• /
• v.17 no.4
• /
• pp.109-116
• /
• 2017

It takes quite amount of time to study a board game because there are many game characters and many state spaces are exist for board games. Therefore, game must do learning long. But, there is weakness with reinforcement learning. On Learning early, the learning speed becomes slow. If there were equal result that both are considered to be best ones during the course of learning stage, Heuristic which utilizes learning of problem area of Jul-Gonu was used to improve the speed of learning. To compare a normal character to an improved one, a board game was created, and then they fought against each other. As a result, improved character's ability was improved on learning speed.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.385-395
• /
• 2005

Long sentence analysis has been a critical problem in machine translation because of high complexity. The methods of intra-sentence segmentation have been proposed to reduce parsing complexity. This paper presents the intra-sentence segmentation method based on maximum entropy probability model to increase the coverage and accuracy of the segmentation. We construct the rules for choosing candidate segmentation positions by a teaming method using the lexical context of the words tagged as segmentation position. We also generate the model that gives probability value to each candidate segmentation positions. The lexical contexts are extracted from the corpus tagged with segmentation positions and are incorporated into the probability model. We construct training data using the sentences from Wall Street Journal and experiment the intra-sentence segmentation on the sentences from four different domains. The experiments show about $88\%$ accuracy and about $98\%$ coverage of the segmentation. Also, the proposed method results in parsing efficiency improvement by 4.8 times in speed and 3.6 times in space.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.13 no.5
• /
• pp.313-322
• /
• 2020

A large volume of data will improve the robustness of deep learning models and avoid overfitting problems. In automatic tongue segmentation, the availability of annotated tongue images is often limited because of the difficulty of collecting and labeling the tongue image datasets in reality. Data augmentation can expand the training dataset and increase the diversity of training data by using label-preserving transformations without collecting new data. In this paper, augmented tongue image datasets were developed using seven augmentation techniques such as image cropping, rotation, flipping, color transformations. Performance of the data augmentation techniques were studied using state-of-the-art transfer learning models, for instance, InceptionV3, EfficientNet, ResNet, DenseNet and etc. Our results show that geometric transformations can lead to more performance gains than color transformations and the segmentation accuracy can be increased by 5% to 20% compared with no augmentation. Furthermore, a random linear combination of geometric and color transformations augmentation dataset gives the superior segmentation performance than all other datasets and results in a better accuracy of 94.98% with InceptionV3 models.

• Communications of Mathematical Education
• /
• v.36 no.1
• /
• pp.39-57
• /
• 2022

This study aims to design an integral instruction method that follows the Abstraction in Context (AiC) framework proposed by Hershkowitz, Schwarz, and Dreyfus to help students in acquiring in-depth understanding of the relationship between indefinite integrals and definite integrals and to analyze how the students' understanding improved as a result. To this end, we implemented lessons according to the integral instruction method designed for eight 11th grade students in a science high school. We recorded and analyzed data from graded student worksheets and transcripts of classroom recordings. Results show that students comprehend three knowledge elements regarding relationship between indefinite integral and definite integral: the instantaneous rate of change of accumulation function, the calculation of a definite integral through an indefinite integral, and The determination of indefinite integral by the accumulation function. The findings suggest that the AiC framework is useful for designing didactical activities for conceptual learning, and the accumulation function can serve as a basis for teaching the three knowledge elements regarding relationship between indefinite integral and definite integral.

• Journal of KIISE:Software and Applications
• /
• v.35 no.9
• /
• pp.547-553
• /
• 2008

We propose a two-stage document layout segmentation method. At the first stage, as top-down segmentation, morphological distance map algorithm extracts a collection of rectangular regions from a given input image. This preliminary result from the first stage is employed as input parameters for the process of next stage. At the second stage, a machine-learning algorithm is adopted RBF network, one of neural networks based on statistical model, is selected. In order for constructing the hidden layer of RBF network, a data clustering technique bared on the self-organizing property of Kohonen network is utilized. We present a result showing that the supervised neural network, trained by 300 number of sample data, improves the preliminary results of the first stage.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.347-348
• /
• 2024

본 논문에서는 철도 구성요소 모니터링을 위한 효율적인 객체 분할 기법으로 사전학습된 SegFormer 모델의 적용을 제안하고, 객체 분할을 위해 보편적으로 사용되는 U-Net 모델과의 성능 비교 분석을 진행하였다. 철도의 주요 구성요소인 선로, 침목, 고정 장치, 배경을 분할할 수 있도록 라벨링된 데이터셋을 학습에 사용하였다. SegFormer 모델이 대조군인 U-Net보다 성능이 Jaccard Score 기준 5.29% 향상됨에 따라 Vision Transformer 기반의 모델이 기존 CNN 기반 모델의 이미지의 전역적인 문맥을 파악하기 상대적으로 어렵다는 한계를 극복하고, 철도 구성요소 객체 분할에 더욱 효율적인 모델임을 확인한다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.1
• /
• pp.85-92
• /
• 2020

Deep learning shows outstanding performance in image and video analysis, such as object classification, object detection and semantic segmentation. In this paper, it is analyzed that the performances of deep learning models can be affected by characteristics of train dataset. It is proposed as a method for selecting activation function and optimization algorithm of deep learning to classify facial expression. Classification performances are compared and analyzed by applying various algorithms of each component of deep learning model for CK+, MMI, and KDEF datasets. As results of simulation, it is shown that genetic algorithm can be an effective solution for optimizing components of deep learning model.

• Journal of KIISE:Software and Applications
• /
• v.31 no.10
• /
• pp.1259-1265
• /
• 2004

In many applications, producing labeled data is costly and time consuming while an enormous amount of unlabeled data is available with little cost. Therefore, it is natural to ask whether we can take advantage of these unlabeled data in classification teaming. In machine learning literature, the co-training method has been widely used for this purpose. However, the current co-training method requires the entire features to be split into two independent sets. Therefore, in this paper, we improved the current co-training method in a number of ways, and proposed a new co-training method which do not need the feature split. Experimental results show that our proposed method can significantly improve the performance of the current co-training algorithm.