• Korean Journal of Remote Sensing
• /
• v.39 no.5_4
• /
• pp.1111-1123
• /
• 2023

In recent times, semantic image segmentation methods using deep learning models have been widely used for monitoring changes in surface attributes using remote sensing imagery. To enhance the performance of various UNet-based deep learning models, including the prominent UNet model, it is imperative to have a sufficiently large training dataset. However, enlarging the training dataset not only escalates the hardware requirements for processing but also significantly increases the time required for training. To address these issues, transfer learning is used as an effective approach, enabling performance improvement of models even in the absence of massive training datasets. In this paper we present three transfer learning models, UNet-ResNet50, UNet-VGG19, and CBAM-DRUNet-VGG19, which are combined with the representative pretrained models of VGG19 model and ResNet50 model. We applied these models to building extraction tasks and analyzed the accuracy improvements resulting from the application of transfer learning. Considering the substantial impact of learning rate on the performance of deep learning models, we also analyzed performance variations of each model based on different learning rate settings. We employed three datasets, namely Kompsat-3A dataset, WHU dataset, and INRIA dataset for evaluating the performance of building extraction results. The average accuracy improvements for the three dataset types, in comparison to the UNet model, were 5.1% for the UNet-ResNet50 model, while both UNet-VGG19 and CBAM-DRUNet-VGG19 models achieved a 7.2% improvement.

• School Mathematics
• /
• v.8 no.2
• /
• pp.161-176
• /
• 2006

Some adequate programs for mathematising are necessary to pre-service mathematics teachers, if they can guide their prospective students in secondary school to make a mathematising. They should be used to mathematising. In this paper, mathematising teaching units for the inquiry into number partition models with constant differences are designed for this purpose. They guide a series of process to make nooumenon for organizing phainomenon which is organized already through number partition model. Especially the new nooumenon and the process of obtaining it are discussed. But it is restricted when the numbers for partitioning are natural numbers, and elements and their differences are integers. Through these teaching units, pre-service mathematics teachers can experience and practice secondary mathematising, as they go through the procedures which are similar with those of mathematicians making theorems.

• Proceedings of the Korean Society of Soil and Groundwater Environment Conference
• /
• 2003.09a
• /
• pp.493-496
• /
• 2003

이 연구는 우리나라 균열암반 대수층의 수리적 특성을 해석ㆍ평가하기 위하여 양수시험 해석해(Theis, 1935; Cooper-Jacob, 1946; Papadopulos-Cooper, 1967; Hantush, 1962a,b; Moench, 1985; Hantush-Jacob, 1955) 및 일반화 방사상 유동 모델을 이용하여 균열암반 대수층(화강암, 화산암, 변성암, 백악기퇴적암, 제3기 퇴적암에 굴착된 100개 조사공)에서 수행되어진 양수시험으로부터 얻은 122개의 양수시험자료(수위강하 자료)를 분석하였다. AQTESOLV 전산프로그램을 이용한 양수시험자료 분석에 의하면, 122개 자료중 86개(71%)의 자료들이 이 연구에 사용된 해석해와 일치하며, 양수시험자료 해석해 중에 누수(leaky) 및 경계조건(boundary condition)을 고려한 해석해들이 53개(43%)로 가장 많이 나타났다. 그러므로, 양수시험자료의 해석은 균열암반 대수층의 수리지질학적 특성에 적합한 개념모델의 설정이 중요하다. 일반화 방사상 유동(GRF)모델을 적용해보면, 122개의 자료중 77개(63%)의 자료들이 Barker(1988)의 표준곡선에 의한 차원(1.1차원-2.9차원)을 보여준다. 이중 44.2%에 해당하는 39개 자료가 1.1차원과 1.9차원 사이의 분할 유동차원을 보여주는 반면에 26개(6.5%)만이 Theis 이론에 맞는 2차원의 방사상 흐름을 보여주며, 38개(49.3%)는 2.1차원에서 2.9차원에 속한다. 따라서 우리나라 균열암반 대수층에서 지하수 유동은 대부분 분할차원의 유동을 보여주는 것으로 평가된다.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.27 no.1
• /
• pp.30-36
• /
• 2023

In this study, an image segmentation model that can evaluate surface porosity based on concrete surface images was derived. Three types of concrete specimens with different water-cement ratios (w/c = 54, 35, and 30%) were prepared, and 2,729 surface images were obtained using an optical microscope. Benchmarking tests, parameter optimization, and final model derivation were performed using the surface images, and an image segmentation model with 97% verification accuracy was obtained. The model was verified by comparing the porosity obtained from the model and X-Ray Microscope (XRM). The model provided similar porosity to that of XRM for the specimens with a high water-cement ratio, but tended to give lower porosity for specimens with a low water-cement ratio.

• The KIPS Transactions:PartD
• /
• v.15D no.1
• /
• pp.87-98
• /
• 2008

In company with the demand of powerful processing units for embedded systems, the method to develop embedded software is also required to support the demand in new approach. In order to improve the resource utilization and system performance, software modeling techniques have to consider the features of hardware architecture. This paper proposes a partitioning technique of UML-based software models, which focus the generation of the allocatable software components into multiprocessor architecture. Our partitioning technique, at first, transforms UML models to CBCFGs(Constraint-Based Control Flow Graphs), and then slices the CBCFGs with consideration of parallelism and data dependency. We believe that our proposition gives practical applicability in the areas of platform specific modeling and performance estimation in model-driven embedded software development.

• Journal of KIISE:Software and Applications
• /
• v.27 no.8
• /
• pp.795-803
• /
• 2000

This paper presents a shared data decomposition model for improving concurrency in multi-user, distributed software developments. In our model, the target software system is decomposed into the independent components based on project roles to be distributed over clients. The distributed components are decomposed into view objects and core objects to replicate only view objects in a distributed collaboration session. The core objects are kept in only one client and the locking is used to prevent inconsistencies. The grain size of a lock is a role instead of a class which is commonly used as the locking granularity in the existing systems. The experimental result shows that our model reduces response time by 12${\sim}$18% and gives good scalability.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.5
• /
• pp.71-76
• /
• 2023

In this paper, we propose a fuzzy U-Net, a fuzzy deep learning model that applies fuzzy logic to improve performance in image segmentation using deep learning. Fuzzy modules using fuzzy logic were combined with U-Net, a deep learning model that showed excellent performance in image segmentation, and various types of fuzzy modules were simulated. The fuzzy module of the proposed deep learning model learns intrinsic and complex rules between feature maps of images and corresponding segmentation results. To this end, the superiority of the proposed method was demonstrated by applying it to dental CBCT data. As a result of the simulation, it can be seen that the performance of the ADD-RELU fuzzy module structure of the model using the addition skip connection in the proposed fuzzy U-Net is 0.7928 for the test dataset and the best.

• KIPS Transactions on Computer and Communication Systems
• /
• v.9 no.2
• /
• pp.37-44
• /
• 2020

In this paper, we proposed that a fully automatic multi-class whole heart segmentation algorithm using deep learning. The proposed method is based on U-Net architecture which consist of recurrent convolutional block, residual multi-dilated convolutional block. The evaluation was accomplished by comparing automated analysis results of the test dataset to the manual assessment. We obtained the average DSC of 96.88%, precision of 95.60%, and recall of 97.00% with CT images. We were able to observe and analyze after visualizing segmented images using three-dimensional volume rendering method. Our experiment results show that proposed method effectively performed to segment in various heart structures. We expected that our method can help doctors and radiologist to make image reading and clinical decision.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.1
• /
• pp.40-48
• /
• 2007

This paper presents a novel texture segmentation method using multilayer perceptron (MLP) networks and Markov random fields in multiscale Bayesian framework. Multiscale wavelet coefficients are used as input for the neural networks. The output of the neural network is modeled as a posterior probability. Texture classification at each scale is performed by the posterior probabilities from MLP networks and MAP (maximum a posterior) classification. Then, in order to obtain the more improved segmentation result at the finest scale, our proposed method fuses the multiscale MAP classifications sequentially from coarse to fine scales. This process is done by computing the MAP classification given the classification at one scale and a priori knowledge regarding contextual information which is extracted from the adjacent coarser scale classification. In this fusion process, the MRF (Markov random field) prior distribution and Gibbs sampler are used, where the MRF model serves as the smoothness constraint and the Gibbs sampler acts as the MAP classifier. The proposed segmentation method shows better performance than texture segmentation using the HMT (Hidden Markov trees) model and HMTseg.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.11
• /
• pp.1096-1100
• /
• 2010

In this paper, we propose an automatic segmentation of the meniscus based on active shape model using interpolated shape information in MR images. First, the statistical shape model of meniscus is constructed to reflect the shape variation in the training set. Second, the generation technique of interpolated shape information by using the weight according to shape similarity is proposed to robustly segment the meniscus with large variation. Finally, the automatic meniscus segmentation is performed through the active shape model fitting. For the evaluation of our method, we performed the visual inspection, accuracy measure and processing time. For accuracy evaluation, the average distance difference between automatic segmentation and semi-automatic segmentation are calculated and visualized by color-coded mapping. Experimental results show that the average distance difference was $0.54{\pm}0.16mm$ in medial meniscus and $0.73{\pm}0.39mm$ in lateral meniscus. The total processing time was 4.87 seconds on average.