• Journal of Internet Computing and Services
• /
• v.24 no.4
• /
• pp.85-92
• /
• 2023

Deep learning applications are increasingly being leveraged for disease detection tasks in medical imaging modalities such as X-ray, Computed Tomography (CT), and Magnetic Resonance Imaging (MRI). Most data-centric deep learning challenges necessitate the use of supervised learning methodologies to attain high accuracy and to facilitate performance evaluation through comparison with the ground truth. Supervised learning mandates a substantial amount of image and label sets, however, procuring an adequate volume of medical imaging data for training is a formidable task. Various data augmentation strategies can mitigate the underfitting issue inherent in supervised learning-based models that are trained on limited medical image and label sets. This research investigates the enhancement of a deep learning-based rib fracture segmentation model and the efficacy of data augmentation techniques such as left-right flipping, rotation, and scaling. Augmented dataset with L/R flipping and rotations(30°, 60°) increased model performance, however, dataset with rotation(90°) and ⨯0.5 rescaling decreased model performance. This indicates the usage of appropriate data augmentation methods depending on datasets and tasks.

• Journal of Korea Multimedia Society
• /
• v.21 no.2
• /
• pp.98-107
• /
• 2018

In this paper, we propose a method to classify the skeletal maturity with a small amount of hand wrist X-ray image using deep learning-based meta-learning. General deep-learning techniques require large amounts of data, but in many cases, these data sets are not available for practical application. Lack of learning data is usually solved through transfer learning using pre-trained models with large data sets. However, transfer learning performance may be degraded due to over fitting for unknown new task with small data, which results in poor generalization capability. In addition, medical images require high cost resources such as a professional manpower and mcuh time to obtain labeled data. Therefore, in this paper, we use meta-learning that can classify using only a small amount of new data by pre-trained models trained with various learning tasks. First, we train the meta-model by using a separate data set composed of various learning tasks. The network learns to classify the bone maturity using the bone maturity data composed of the radiographs of the wrist. Then, we compare the results of the classification using the conventional learning algorithm with the results of the meta learning by the same number of learning data sets.

• Convergence Security Journal
• /
• v.8 no.4
• /
• pp.161-166
• /
• 2008

This paper proposes a new approach to align 3D data sets by using curvatures of feature surface. We use the Gaussian curvatures and the covariance matrix which imply the physical characteristics of the model to achieve registration of unaligned 3D data sets. First, the physical characteristics of local area are obtained by the Gaussian curvature. And the camera position of 3D range finder system is calculated from by using the projection matrix between 3D data set and 2D image. Then, the physical characteristics of whole area are obtained by the covariance matrix of the model. The corresponding points can be found in the overlapping region with the cross-projection method and it concentrates by removed points of self-occlusion. By the repeatedly the process discussed above, we finally find corrected points of overlapping region and get the optimized registration result.

• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.148-153
• /
• 1999

Baikdu-san was a very active volcano during the Cenozoic era and is believed to be formed in late Cenozoic era. Recently it was also reported that there was a major eruption in or around 1002 A.D. and there are evidences which indicate that it is still an active volcano and a potential volcanic hazard. Remote sensing techniques have been widely used to monitor various natural hazards, including volcanic hazards. However, during an active volcanic eruption, volcanic ash can basically cover the sky and often blocks the solar radiation preventing any use of optical sensors. Synthetic aperture radar(SAR) is an ideal tool to monitor the volcanic activities and lava flows, because the wavelength of the microwave signal is considerably longer that the average volcanic ash particle size. In this study we have utilized several sets of SAR data to evaluate the utility of the space-borne SAR system. The data sets include JERS-1(L-band) SAR, and RADARSAT(C-band) data which included both standard mode and the ScanSAR mode data sets. We also utilized several sets of auxiliary data such as local geological maps and JERS-1 OPS data. The routine preprocessing and image processing steps were applied to these data sets before any attempts of classifying and mapping surface geological features. Although we computed sigma nought ($\sigma$$^{0}$) values far the standard mode RADARSAT data, the utility of sigma nought image was minimal in this study. Application of various types of classification algorithms to identify and map several stages of volcanic flows was not very successful. Although this research is still in progress, the following preliminary conclusions could be made: (1) sigma nought (RADARSAT standard mode data) and DN (JERS-1 SAR and RADARSAT ScanSAR data) have limited usefulness for distinguishing early basalt lava flows from late trachyte flows or later trachyte flows from the old basement granitic rocks around Baikdu-san volcano, (2) surface geological structure features such as several faults and volcanic lava flow channels can easily be identified and mapped, and (3) routine application of unsupervised classification methods cannot be used for mapping any types of surface lava flow patterns.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.29 no.4
• /
• pp.429-438
• /
• 2011

LiDAR technology is a combination of laser ranging, satellite positioning technology and digital image technology for study and determination with high accuracy of the true earth surface features in 3 D. Laser scanning data is typically a points cloud on the ground, including coordinates, altitude and intensity of laser from the object on the ground to the sensor (Wehr & Lohr, 1999). Data from laser scanning can produce products such as digital elevation model (DEM), digital surface model (DSM) and the intensity data. In Vietnam, the LiDAR technology has been applied since 2005. However, the application of LiDAR in Vietnam is mostly for topological mapping and DEM establishment using point cloud 3D coordinate. In this study, another application of LiDAR data are present. The study use the intensity image combine with some other data sets (elevation data, Panchromatic image, RGB image) in Bacgiang City to perform land cover classification using neural network method. The results show that it is possible to obtain land cover classes from LiDAR data. However, the highest accurate classification can be obtained using LiDAR data with other data set and the neural network classification is more appropriate approach to conventional method such as maximum likelyhood classification.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.30B no.2
• /
• pp.69-77
• /
• 1993

Automatic image recognition system is essential for a better man-to machine interaction. Because of the noise and deformation due to the sensor operation, it is not simple to build an image recognition system even for the fixed images. In this paper neural network which has been reported to be adequate for pattern recognition task is applied to the fixed and variational(rotation, size, position variation for the fixed image)recognition with a hope that the problems of conventional pattern recognition techniques are overcome. At fixed image recognition system. ART model is trained with face images obtained by camera. When recognizing an matching score. In the test when wigilance level 0.6 - 0.8 the system has achievel 100% correct face recognition rate. In the variational image recognition system, 65 invariant moment features sets are taken from thirteen persons. 39 data are taken to train multi-layer perceptron and other 26 data used for testing. The result shows 92.5% recognition rate.

• Proceedings of the Korean Institute of Communication Sciences Conference
• /
• 2004.07a
• /
• pp.219-219
• /
• 2004

• Smart Media Journal
• /
• v.8 no.2
• /
• pp.46-57
• /
• 2019

In recent research, deep learning-based methods have achieved state-of-the-art performance in various computer vision tasks. However, these methods are commonly supervised, and require huge amounts of annotated data to train. Acquisition of data demands an additional costly effort, particularly for the tasks where it becomes challenging to obtain large amounts of data considering the time constraints and the requirement of professional human diligence. In this paper, we present a data level synthetic sampling solution to learn from small and imbalanced data sets using Generative Adversarial Networks (GANs). The reason for using GANs are the challenges posed in various fields to manage with the small datasets and fluctuating amounts of samples per class. As a result, we present an approach that can improve learning with respect to data distributions, reducing the partiality introduced by class imbalance and hence shifting the classification decision boundary towards more accurate results. Our novel method is demonstrated on a small dataset of 2789 tomato plant disease images, highly corrupted with class imbalance in 9 disease categories. Moreover, we evaluate our results in terms of different metrics and compare the quality of these results for distinct classes.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.4B
• /
• pp.743-754
• /
• 2000

In this paper, we propose wavelet transform image compression method using shuffling and bit plane correlation. Proposed method is that original image decompose into multiresolutions using biorthogonal wavelet transform with linear phase response property and decomposed subbands are classified by maximum classification gain. And classified data sets in each subband are quantized using arbitrary set optimum bit allocation method. Quantized data sets in each subband are shuffled and context based bit plane arithmetic encoded .In context based bit plane arithmetic encoding, the context for each subband is not assigned uniformly, but assigned according to maximum correlation direction. Our results are comparable, or superior for some images at low rates, to published state-of-the-art coders.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.1084-1086
• /
• 2003

The aim of this study is to assess the integrated use of different features extracted from spaceborne interferometric synthetic aperture radar (InSAR) data and optical data for land cover classification. Special attention is given to the discriminatory characteristics of the features derived from the multisource data sets. For the evaluation of the features , the statistical maximum likelihood decision rule and neural network classification are used and the results are compared. The performance of each method was evaluated by measuring the overall accuracy. In all cases, the performance of the first method was better than the performance of the latter one. Overall, the research indicated that multisource data sets containing different information about backscattering and reflecting properties of the selected classes of objects can significantly improve the classification of land cover types.