• Journal of KIISE
• /
• v.41 no.11
• /
• pp.921-926
• /
• 2014

Face detection refers to the process extracting facial regions in an input image, which can improve speed and accuracy of recognition or authorization system, and has diverse applicability. Since conventional works have tried to detect faces based on the whole shape of faces, its detection performance can be degraded by occlusion made with accessories or parts of body. In this paper we propose a method combining local feature descriptors and probability modeling in order to detect partially occluded face effectively. In training stage, we represent an image as a set of local feature descriptors and estimate a statistical model for normal faces. When the test image is given, we find a region that is most similar to face using our face model constructed in training stage. According to experimental results with benchmark data set, we confirmed the effect of proposed method on detecting partially occluded face.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.5
• /
• pp.697-704
• /
• 2018

Since the face in image content corresponds to individual information that can distinguish a specific person from other people, it is important to accurately detect faces not hidden in an image. In this paper, we propose a method to accurately detect a face from input images using a deep learning algorithm, which is one of the machine learning methods. In the proposed method, image input via the red-green-blue (RGB) color model is first changed to the luminance-chroma: blue-chroma: red-chroma ($YC_bC_r$) color model; then, other regions are removed using the learned skin color model, and only the skin regions are segmented. A CNN model-based deep learning algorithm is then applied to robustly detect only the face region from the input image. Experimental results show that the proposed method more efficiently segments facial regions from input images. The proposed face area-detection method is expected to be useful in practical applications related to multimedia and shape recognition.

• Journal of the Korean Association of Oral and Maxillofacial Surgeons
• /
• v.33 no.2
• /
• pp.81-93
• /
• 2007

Purpose: Adenoid cystic carcinoma (ACC) is a relatively rare tumor that arises in glandular tissues of the head and neck region and sometimes has a protracted clinical course with perineural invasion and delayed onset of distant lung metastasis. Treatment failure of salivary ACC is most often associated with perineural and hematogenous tumor spread. However, very little has been known about the cellular and molecular mechanisms of perineural invasion and hematogenous distant metastasis of parotid ACC. This study was designed to develop an orthotopic tumor model of parotid adenoid cystic carcinoma in athymic nude mice. Experimental Design: A melanoma cell line was injected into the parotid gland of athymic mice to determine whether such implantation was technically feasible. A parotid ACC cell line was then injected into the parotid gland or the subcutaneous tissue of athymic mice at various concentrations of tumor cells, and the mice were thereafter followed for development of tumor nodule. The tumors were examined histopathologically for perineural invasion or regional or distant lung metastasis. We used an oral squmous cell carcinoma cell line as control. Results: Implantation of tumor(melanoma) cell suspension into the parotid gland of nude mice was technically feasible and resulted in the formation of parotid tumors. A parotid ACC cell line, ACC3 showed no significantly higher tumorigenicity, but showed significantly higher lung metastatic potential in the parotid gland than in the subcutis. In contrast, mucosal squmous cell carcinoma cell line doesn’t show significantly higher lung metastatic potential in the parotid gland than in the subcutis. The ACC tumor established in the parotid gland seemed to demonstrate perineural invasion of facial nerve, needs further study. Conclusion: An orthotopic tumor model of salivary ACC in athymic nude mice was successfully developed that closely recapitulates the clinical situations of human salivary ACC. This model should facilitate the understanding of the cellular and molecular mechanisms of tumorigenisis and metastasis of salivary ACC and aid in the development of targeted molecular therapies of salivary ACC.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.16 no.6
• /
• pp.759-765
• /
• 2006

In this paper, it is proposed an effective face expression recognition LDA mixed mode using a triangularity membership fuzzy function and wavelet basis. The proposal algorithm gets performs the optimal image, fuzzy wavelet algorithm and Expression recognition is consisted of face characteristic detection step and face Expression recognition step. This paper could applied to the PCA and LDA in using some simple strategies and also compares and analyzes the performance of the LDA mixed model which is combined and the facial expression recognition based on PCA and LDA. The LDA mixed model is represented by the PCA and the LDA approaches. And then we calculate the distance of vectors dPCA, dLDA from all fates in the database. Last, the two vectors are combined according to a given combination rule and the final decision is made by NNPC. In a result, we could showed the superior the LDA mixed model can be than the conventional algorithm.

• Journal of Korea Multimedia Society
• /
• v.25 no.10
• /
• pp.1375-1385
• /
• 2022

In this paper, we propose a semi-supervised domain adaptation solution to deal with practical face recognition (FR) scenarios where a single face image for each target identity (to be recognized) is only available in the training phase. Main goal of the proposed method is to reduce the discrepancy between the target and the source domain face images, which ultimately improves FR performances. The proposed method is based on the Domain Adatation network (DAN) using an MMD loss function to reduce the discrepancy between domains. In order to train more effectively, we develop a novel loss function learning strategy in which MMD loss and cross-entropy loss functions are adopted by using different weights according to the progress of each epoch during the learning. The proposed weight adoptation focuses on the training of the source domain in the initial learning phase to learn facial feature information such as eyes, nose, and mouth. After the initial learning is completed, the resulting feature information is used to training a deep network using the target domain images. To evaluate the effectiveness of the proposed method, FR performances were evaluated with pretrained model trained only with CASIA-webface (source images) and fine-tuned model trained only with FERET's gallery (target images) under the same FR scenarios. The experimental results showed that the proposed semi-supervised domain adaptation can be improved by 24.78% compared to the pre-trained model and 28.42% compared to the fine-tuned model. In addition, the proposed method outperformed other state-of-the-arts domain adaptation approaches by 9.41%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.3
• /
• pp.1100-1118
• /
• 2021

In a wide range of ML applications, the training data contains privacy-sensitive information that should be kept secure. Training the ML systems by privacy-sensitive data makes the ML model inherent to the data. As the structure of the model has been fine-tuned by training data, the model can be abused for accessing the data by the estimation in a reverse process called model inversion attack (MIA). Although, MIA has been applied to shallow neural network models of recognizers in literature and its threat in privacy violation has been approved, in the case of a deep learning (DL) model, its efficiency was under question. It was due to the complexity of a DL model structure, big number of DL model parameters, the huge size of training data, big number of registered users to a DL model and thereof big number of class labels. This research work first analyses the possibility of MIA on a deep learning model of a recognition system, namely a face recognizer. Second, despite the conventional MIA under the white box scenario of having partial access to the users' non-sensitive information in addition to the model structure, the MIA is implemented on a deep face recognition system by just having the model structure and parameters but not any user information. In this aspect, it is under a semi-white box scenario or in other words a gray-box scenario. The experimental results in targeting five registered users of a CNN-based face recognition system approve the possibility of regeneration of users' face images even for a deep model by MIA under a gray box scenario. Although, for some images the evaluation recognition score is low and the generated images are not easily recognizable, but for some other images the score is high and facial features of the targeted identities are observable. The objective and subjective evaluations demonstrate that privacy cyber-attack by MIA on a deep recognition system not only is feasible but also is a serious threat with increasing alert state in the future as there is considerable potential for integration more advanced ML techniques to MIA.

• Journal of Society of Preventive Korean Medicine
• /
• v.21 no.3
• /
• pp.1-10
• /
• 2017

Objectives : The purpose of this study is to investigate the practice patterns of traditional Korean medicine (TKM) doctors and the acceptance of payment model in order to develop a new TKM health insurance payment model linked with TKM clinical practice guidelines (CPGs). Methods : Lumbar herniated intervertebral disc (HIVD) and idiopathic facial palsy (IFP) were selected as a test diseases to develop a new TKM payment model. The level of benefit coverage in the National Health Insurance (NHI) was designed. The survey asked 228 TKM doctors about their practice patterns in HIVD and IFP patients and acceptance of new payment model. Results : Mean of medical cost for treatment of HIVD was 441,000 KW, mean of treatment period ranged from 4.9 to 17.5 weeks, and mean of number of treatment ranged from 14.6 to 50.4 HIVD patients. In the case of IFP, mean of medical cost for treatment of IFP was 468,000 KW, mean of treatment period was at least 4.2 and up to 15.9 weeks and mean of number of treatment ranged from 14.2 to 52 IFP patients. Conclusions : Current study suggests that mixed payment model of per-visit and episode-based model seem to be proper. The model 1 bundles both items which were covered and not covered by NHI in a rational way. The model 2 is based on the development and application of critical pathway. Lastly, model 3 suggests bundling of items covered by current NHI. Acceptance of TKM doctors is expected to be highest in the model 3.

• Archives of Plastic Surgery
• /
• v.34 no.1
• /
• pp.18-23
• /
• 2007

Purpose: Anthropometry can be divided into two methods, direct anthropometry and indirect anthropometry. The most ideal and accurate method is a direct anthropometry. However, it is difficult to measure in the case of children because of poor cooperation, and it lacks re-productivity. Cast model has advantages of three dimensional featuring, inexpensive and easy fabrication. This study is conducted to find out an accuracy of indirect anthropometry on cast model by comparing it with direct anthropometry. Methods: Total 48 cleft lip nasal deformity patients (unilateral, 40; bilateral, 8) were included in this study. Cast models were made before surgery under general anesthesia with alginate impression material and model plaster. Eleven linear measurements among 7 landmarks were taken as direct anthropometry before surgery with Castroviejo spreading caliper. At the same time, indirect anthropometry on cast model was done at the same linear distances as well. Results: Of the total 11 linear measurements, both ala lengths, both columella lengths, nose width, projective distance between facial insertion points of the ala, projective distance between the alar base points, right nostril floor width, and columella width were statistically correlated between indirect anthropometry on cast model and direct anthropometry. However, the nasal tip protrusion and the left nostril floor width were not statistically correlated. Conclusion: Accuracy of indirect anthropometry on cast model can be influenced by cast model fabrication techniques and correct identification of landmarks. Nasal tip protrusion could be reduced by compression of the nasal tip in the process of cast model fabrication and nostril floor width can be varied by muscle relaxation of anesthetics and incorrect identification of subalare in cleft lip nasal deformity. If sufficient care is taken to make cast model and to define landmarks exactly, indirect anthropometry on cast model can be a reliable method as direct anthropometry.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.3
• /
• pp.316-322
• /
• 2014

A movie recommendation system is proposed to learn a preference model of a viewer by using multimodal features of a video content and their evoked implicit responses of the viewer in synchronized manner. In this system, facial expression, body posture, and physiological signals are measured to estimate the affective states of the viewer, in accordance with the stimuli consisting of low-level and affective features from video, audio, and text streams. Experimental results show that it is possible to predict arousal response, which is measured by electrodermal activity, of a viewer from auditory and text features in a video stimuli, for estimating interestingness on the video.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.29 no.4
• /
• pp.309-315
• /
• 2016

Recently, finite element analysis technique has been widely used for structural and mechanical understandings of human body in the dentistry field. This research proposed an effective finite element modeling method based on CT images, and parametric studies were performed for the occlusal simulation. The analyses were performed considering linear material behaviors and nonlinear geometrical effect, and validated with the experimental results. In addition, the skull models with two different molar relations such as Class I and full-CUSP Class II were generated and the analyses were performed using the proposed analytical method. As results, the relationships between the mandibular movement and occlusal force of both two models showed similar tendency in human occlusal force. However, stress was evenly distributed from teeth to facial bone in the skull model with Class I, while stress concentration was appeared in the model with full-CUSP Class II due to the changes of occlusal surfaces of the model.