• Journal of Korean Neurosurgical Society
• /
• v.63 no.3
• /
• pp.386-396
• /
• 2020

Objective : To generate synthetic spine magnetic resonance (MR) images from spine computed tomography (CT) using generative adversarial networks (GANs), as well as to determine the similarities between synthesized and real MR images. Methods : GANs were trained to transform spine CT image slices into spine magnetic resonance T2 weighted (MRT2) axial image slices by combining adversarial loss and voxel-wise loss. Experiments were performed using 280 pairs of lumbar spine CT scans and MRT2 images. The MRT2 images were then synthesized from 15 other spine CT scans. To evaluate whether the synthetic MR images were realistic, two radiologists, two spine surgeons, and two residents blindly classified the real and synthetic MRT2 images. Two experienced radiologists then evaluated the similarities between subdivisions of the real and synthetic MRT2 images. Quantitative analysis of the synthetic MRT2 images was performed using the mean absolute error (MAE) and peak signal-to-noise ratio (PSNR). Results : The mean overall similarity of the synthetic MRT2 images evaluated by radiologists was 80.2%. In the blind classification of the real MRT2 images, the failure rate ranged from 0% to 40%. The MAE value of each image ranged from 13.75 to 34.24 pixels (mean, 21.19 pixels), and the PSNR of each image ranged from 61.96 to 68.16 dB (mean, 64.92 dB). Conclusion : This was the first study to apply GANs to synthesize spine MR images from CT images. Despite the small dataset of 280 pairs, the synthetic MR images were relatively well implemented. Synthesis of medical images using GANs is a new paradigm of artificial intelligence application in medical imaging. We expect that synthesis of MR images from spine CT images using GANs will improve the diagnostic usefulness of CT. To better inform the clinical applications of this technique, further studies are needed involving a large dataset, a variety of pathologies, and other MR sequence of the lumbar spine.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.1
• /
• pp.30-44
• /
• 2022

The advent of deep learning-based algorithms has facilitated researches on target detection from synthetic aperture radar(SAR) imagery. While most of them concentrate on detection tasks for ships with open SAR ship datasets and for aircraft from SAR scenes of airports, there is relatively scarce researches on the detection of SAR ground vehicle targets where several adverse factors such as high false alarm rates, low signal-to-clutter ratios, and multiple targets in close proximity are predicted to degrade the performances. In this paper, a dataset of ground vehicle targets acquired from TerraSAR-X(TSX) satellite SAR images is presented. Then, both detection and instance segmentation are simultaneously carried out on this dataset based on the deep learning-based Mask R-CNN. Finally, this paper shows the future research directions to further improve the performances of detecting the SAR ground vehicle targets.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.465-470
• /
• 2024

The advancement of AI-based image generation technology has resulted in the creation of various images, emphasizing the need for technology capable of accurately discerning them. The amount of generated image data is limited, and to achieve high performance with a limited dataset, this study proposes a model for discriminating generated images using transfer learning. Applying pre-trained models from the ImageNet dataset directly to the CIFAKE input dataset, we reduce training time cost followed by adding three hidden layers and one output layer to fine-tune the model. The modeling results revealed an improvement in the performance of the model when adjusting the final layer. Using transfer learning and then adjusting layers close to the output layer, small image data-related accuracy issues can be reduced and generated images can be classified.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.4
• /
• pp.381-389
• /
• 2022

Detecting partially occluded objects is difficult due to the appearances and shapes of occluders are highly variable. These variabilities lead to challenges of localizing accurate bounding box or classifying objects with visible object parts. To address these problems, we propose a two-stage part-based attention approach for robust object detection under partial occlusion. First, our part attention network(PAN) captures the important object parts and then it is used to generate weighted object features. Based on the weighted features, the re-weighted object features are produced by our reinforced PAN(RPAN). Experiments are performed on our collected military vehicle dataset and synthetic occlusion dataset. Our method outperforms the baselines and demonstrates the robustness of detecting objects under partial occlusion.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3815-3835
• /
• 2022

Density-based outlier detection is one of the hot issues in data mining. A point is determined as outlier on basis of the density of points near them. The existing density-based detection algorithms have high time complexity, in order to reduce the time complexity, a new outlier detection algorithm DODMD (Density-based Outlier Detection in Multidimensional Datasets) is proposed. Firstly, on the basis of ZH-tree, the concept of micro-cluster is introduced. Each leaf node is regarded as a micro-cluster, and the micro-cluster is calculated to achieve the purpose of batch filtering. In order to obtain n sets of approximate outliers quickly, a greedy method is used to calculate the boundary of LOF and mark the minimum value as LOF_min. Secondly, the outliers can filtered out by LOF_min, the real outliers are calculated, and then the result set is updated to make the boundary closer. Finally, the accuracy and efficiency of DODMD algorithm are verified on real dataset and synthetic dataset respectively.

• Smart Media Journal
• /
• v.12 no.5
• /
• pp.9-16
• /
• 2023

In the recommendation system of the credit card company, it is necessary to understand the customer patterns to predict a customer's next merchant based on their histories. The data we want to model is much more complex and there are various patterns that customers choose. In such a situation, it is necessary to use an effective model that not only shows the relevance of the merchants, but also the relevance of the customers relative to these merchants. The proposed model aims to predict the next merchant for the customer. To improve prediction performance, we propose a novel model, called Customer-based Recommendation Model (CRM), to produce a more efficient representation of customers. For the next merchant recommendation system, we use a synthetic credit card usage dataset, BC'17. To demonstrate the applicability of the proposed model, we also apply it to the next item recommendation with another real-world transaction dataset, IJCAI'16.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.237-250
• /
• 2022

Structural health monitoring (SHM) plays an important role in ensuring the safety and functionality of critical civil infrastructure. In recent years, numerous researchers have conducted studies to develop computer vision and machine learning techniques for SHM purposes, offering the potential to reduce the laborious nature and improve the effectiveness of field inspections. However, high-quality vision data from various types of damaged structures is relatively difficult to obtain, because of the rare occurrence of damaged structures. The lack of data is particularly acute for fatigue crack in steel bridge girder. As a result, the lack of data for training purposes is one of the main issues that hinders wider application of these powerful techniques for SHM. To address this problem, the use of synthetic data is proposed in this article to augment real-world datasets used for training neural networks that can identify fatigue cracks in steel structures. First, random textures representing the surface of steel structures with fatigue cracks are created and mapped onto a 3D graphics model. Subsequently, this model is used to generate synthetic images for various lighting conditions and camera angles. A fully convolutional network is then trained for two cases: (1) using only real-word data, and (2) using both synthetic and real-word data. By employing synthetic data augmentation in the training process, the crack identification performance of the neural network for the test dataset is seen to improve from 35% to 40% and 49% to 62% for intersection over union (IoU) and precision, respectively, demonstrating the efficacy of the proposed approach.

• Journal of Korea Multimedia Society
• /
• v.24 no.10
• /
• pp.1358-1368
• /
• 2021

A research on agricultural automation is a main issues to overcome the shortage of labor in Korea. A sweet persimmon farmers need much time and labors for classifying profitable sweet persimmon and ill profitable products. In this paper, we propose a mixed two-step synthetic neural network model for efficiently classifying sweet persimmon images. In this model, we suggested a surface direction classification model and a quality screening model which constructed from image data sets. Also we studied Class Activation Mapping(CAM) for visualization to easily inspect the quality of the classified products. The proposed mixed two-step model showed high performance compared to the simple binary classification model and the multi-class classification model, and it was possible to easily identify the weak parts of the classification in a dataset.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.3
• /
• pp.639-647
• /
• 2011

In this paper, we introduce a design methodology of data-centroid Radial Basis Function neural networks with extended polynomial function. The two underlying design mechanisms of such networks involve K-means clustering method and Particle Swarm Optimization(PSO). The proposed algorithm is based on K-means clustering method for efficient processing of data and the optimization of model was carried out using PSO. In this paper, as the connection weight of RBF neural networks, we are able to use four types of polynomials such as simplified, linear, quadratic, and modified quadratic. Using K-means clustering, the center values of Gaussian function as activation function are selected. And the PSO-based RBF neural networks results in a structurally optimized structure and comes with a higher level of flexibility than the one encountered in the conventional RBF neural networks. The PSO-based design procedure being applied at each node of RBF neural networks leads to the selection of preferred parameters with specific local characteristics (such as the number of input variables, a specific set of input variables, and the distribution constant value in activation function) available within the RBF neural networks. To evaluate the performance of the proposed data-centroid RBF neural network with extended polynomial function, the model is experimented with using the nonlinear process data(2-Dimensional synthetic data and Mackey-Glass time series process data) and the Machine Learning dataset(NOx emission process data in gas turbine plant, Automobile Miles per Gallon(MPG) data, and Boston housing data). For the characteristic analysis of the given entire dataset with non-linearity as well as the efficient construction and evaluation of the dynamic network model, the partition of the given entire dataset distinguishes between two cases of Division I(training dataset and testing dataset) and Division II(training dataset, validation dataset, and testing dataset). A comparative analysis shows that the proposed RBF neural networks produces model with higher accuracy as well as more superb predictive capability than other intelligent models presented previously.

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.151-160
• /
• 2012

Most of the distributed high-dimensional indexing structures provide a reasonable search performance especially when the dataset is uniformly distributed. However, in case when the dataset is clustered or skewed, the search performances gradually degrade as compared with the uniformly distributed dataset. We propose a method of improving the k-nearest neighbor search performance for the distributed vector approximation-tree based on the strongly clustered or skewed dataset. The basic idea is to compute volumes of the leaf nodes on the top-tree of a distributed vector approximation-tree and to assign different number of bits to them in order to assure an identification performance of vector approximation. In other words, it can be done by assigning more bits to the high-density clusters. We conducted experiments to compare the search performance with the distributed hybrid spill-tree and distributed vector approximation-tree by using the synthetic and real data sets. The experimental results show that our proposed scheme provides consistent results with significant performance improvements of the distributed vector approximation-tree for strongly clustered or skewed datasets.