• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.568-570
• /
• 2023

The problem of classifying of age, gender, and race images still poses challenges. Despite deep and machine learning strides, convolutional neural networks (CNNs) remain pivotal in addressing these issues. This paper introduces a novel CNN-based approach for accurate and efficient age, gender, and race classification. Leveraging CNNs with residual blocks, our method enhances learning while minimizing computational complexity. The model effectively captures low-level and high-level features, yielding improved classification accuracy. Evaluation of the diverse 'fair face' dataset shows our model achieving 56.3%, 94.6%, and 58.4% accuracy for age, gender, and race, respectively.

• Current Optics and Photonics
• /
• v.7 no.5
• /
• pp.511-517
• /
• 2023

Pipelines play an important role in urban water supply and drainage, oil and gas transmission, etc. This paper presents a technique for pattern recognition of fiber optic vibration signals collected by a distributed vibration sensing (DVS) system using a deep learning residual network (ResNet). The optical fiber is laid on the pipeline, and the signal is collected by the DVS system and converted into a 64 × 64 single-channel grayscale image. The grayscale image is input into the ResNet to extract features, and finally the K-nearest-neighbors (KNN) algorithm is used to achieve the classification and recognition of pipeline damage.

• The Journal of Korea Robotics Society
• /
• v.19 no.2
• /
• pp.196-202
• /
• 2024

The ability to understand given environments and plan a sequence of actions leading to goal state is crucial for personal service robots. With recent advancements in deep learning, numerous studies have proposed methods for state representation in planning. However, previous works lack explicit information about relationships between objects when the state observation is converted to a single visual embedding containing all state information. In this paper, we introduce graph-based state representation that incorporates both object and relationship features. To leverage these advantages in addressing the task planning problem, we propose a Graph Neural Network (GNN)-based subgoal prediction model. This model can extract rich information about object and their interconnected relationships from given state graph. Moreover, a search-based algorithm is integrated with pre-trained subgoal prediction model and state transition module to explore diverse states and find proper sequence of subgoals. The proposed method is trained with synthetic task dataset collected in simulation environment, demonstrating a higher success rate with fewer additional searches compared to baseline methods.

• International journal of advanced smart convergence
• /
• v.13 no.1
• /
• pp.48-56
• /
• 2024

The application of artificial intelligence (AI) is becoming increasingly important to coping with air pollution. AI is effective in coping with it in various ways including air pollution forecasting, monitoring, and control, which is attracting a lot of attention. This attention has created high need for analyzing studies on AI and air pollution. To contribute for satisfying it, this study performed bibliometric analyses on the studies on AI and air pollution from 2012 to 2022 using the Web of Science database. This study analyzed them in various aspects such as the trend in the number of articles, the trend in the number of citations, the top 10 countries of origin, the top 10 research organizations, the top 10 research funding agencies, the top 10 journals, the top 10 articles in terms of total citations, and the distribution by languages. This study not only reports the bibliometric analysis results but also reveals the eight distinct features in the research steam in studies on AI and air pollution, identified from the bibliometric analysis results. They are expected to make a useful contribution for understanding the research stream in AI and air pollution.

• Journal of Korean Neurosurgical Society
• /
• v.67 no.3
• /
• pp.299-307
• /
• 2024

Cerebral cavernous malformation (CCM) is a vascular anomaly commonly found in children and young adults. Common clinical presentations of pediatric patients with CCMs include headache, focal neurological deficits, and seizures. Approximately 40% of pediatric patients are asymptomatic. Understanding the natural history of CCM is crucial and hemorrhagic rates are higher in patients with an initial hemorrhagic presentation, whereas it is low in asymptomatic patients. There is a phenomenon known as temporal clustering in which a higher frequency of symptomatic hemorrhages occurs within a few years following the initial hemorrhagic event. Surgical resection remains the mainstay of treatment for pediatric CCMs. Excision of a hemosiderin-laden rim is controversial regarding its impact on epilepsy outcomes. Stereotactic radiosurgery is an alternative treatment, especially for deep-seated CCMs, but its true efficacy needs to be verified in a clinical trial.

• Clinical Endoscopy
• /
• v.57 no.2
• /
• pp.175-180
• /
• 2024

Therapeutic endoscopic ultrasonography (EUS) procedures using the forward-viewing convex EUS (FV-EUS) have been reviewed based on the articles reported to date. The earliest reported procedure is the drainage of pancreatic pseudocysts using FV-EUS. However, the study on drainage of pancreatic pseudocysts focused on showing that drainage is possible with FV-EUS rather than leveraging its features. Subsequently, studies describing the characteristics of FV-EUS have been reported. By using FV-EUS in EUS-guided choledochoduodenostomy, double punctures in the gastrointestinal tract can be avoided. In postoperative modified anatomical cases, using the endoscopic function of FV-EUS, procedures such as bile duct drainage from anastomosis, pancreatic duct drainage from the afferent limb, and abscess drainage from the digestive tract have been reported. When a perpendicular puncture to the gastrointestinal tract is required or when there is a need to insert the endoscope deep into the gastrointestinal tract, FV-EUS is considered among the options.

• Journal of the Korean earth science society
• /
• v.45 no.4
• /
• pp.279-291
• /
• 2024

We explored the effect of galaxy-galaxy interaction on the FIR-radio correlation of star-forming galaxies by comparing the q_FIR parameter distribution between interacting and non-interacting galaxies. Our sample galaxies were selected from the SDSS Stripe 82 region, where relatively deep optical images are available in addition to ancillary FIR and radio data. The q_FIR values were 2.73±0.49 and 2.53±0.90 for interacting and non-interacting galaxies, respectively. The t-test results indicated that the difference in q_FIR values between the two categories is not statistically significant. Our findings align with those of previous studies suggesting that either FIR excess or radio excess occurs only transiently during brief timescales in the merger stages, rather than persisting throughout the majority of merger events identified by features such as tidal tails or double nuclei.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.3
• /
• pp.27-32
• /
• 2024

Reliable and fine-grained musical metadata are required for efficient search of rapidly increasing music files. In particular, since the primary motive for listening to music is its emotional effect, diversion, and the memories it awakens, emotion classification along with genre classification of music is crucial. In this paper, as an initial approach towards a "ground-truth" dataset for music emotion and genre classification, we elaborately generated a music corpus through labeling of a large number of ordinary people. In order to verify the suitability of the dataset through the classification results, we extracted features according to MPEG-7 audio standard and applied different machine learning models based on statistics and deep neural network to automatically classify the dataset. By using standard hyperparameter setting, we reached an accuracy of 93% for genre classification and 80% for emotion classification, and believe that our dataset can be used as a meaningful comparative dataset in this research field.

• Journal of the Korean Society of Hazard Mitigation
• /
• v.4 no.4 s.15
• /
• pp.43-54
• /
• 2004

The broadband receiver functions are developed from teleseismic P waveforms recorded at Wonju(KSRS), Inchon(IRIS), and Pohang(PHN), and are analyzed to examine the crustal structure beneath these stations. The teleseismic receiver functions are inverted in the time domain of the vertical P wave velocity structures beneath the stations. Clear P-to-S converted phases from the Moho interface are observed in teleseismic seismograms recorded at these stations. The crustal velocity structures beneath the stations are estimated by using the receiver function inversion method(Ammon et al., 1990). The general features of inversion results are as follows: (1) For the Inchon station, the Conrad discontinuity exists at 17.5 Km(SW) deep and the Moho discontinuity exists at 29.5 Km(NW) and 30.5 Km(SE, SW) deep. (2) The shallow crustal structure beneath Wonju station may be covered with a sedimentary rock of a 3 Km thickness. The average Moho depth is assumed about 33.0 Km, and the Conrad discontinuity may exist at 17.0 Km(NE) and 21.0 Km(NW) deep. (3) For Pohang station, the thickness of shallow sedimentary layer is a 3.0 Km in the direction of NE and NW. The Moho depth is 28.0 Km in the direction of the NE and NW. The Conrad discontinuity can be estimated to be existed at 21.0 Km deep for the NE and NW directions.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.3
• /
• pp.133-140
• /
• 2022

We investigate the performance of deep learning-based Korean language models on a task of predicting the score range of Korean essays written by foreign students. We construct a data set containing a total of 304 essays, which include essays discussing the criteria for choosing a job ('job'), conditions of a happy life ('happ'), relationship between money and happiness ('econ'), and definition of success ('succ'). These essays were labeled according to four letter grades (A, B, C, and D), and a total of eleven essay score range prediction experiments were conducted (i.e., five for predicting the score range of 'job' essays, five for predicting the score range of 'happiness' essays, and one for predicting the score range of mixed topic essays). Three deep learning-based Korean language models, KoBERT, KcBERT, and KR-BERT, were fine-tuned using various training data. Moreover, two traditional probabilistic machine learning classifiers, naive Bayes and logistic regression, were also evaluated. Experiment results show that deep learning-based Korean language models performed better than the two traditional classifiers, with KR-BERT performing the best with 55.83% overall average prediction accuracy. A close second was KcBERT (55.77%) followed by KoBERT (54.91%). The performances of naive Bayes and logistic regression classifiers were 52.52% and 50.28% respectively. Due to the scarcity of training data and the imbalance in class distribution, the overall prediction performance was not high for all classifiers. Moreover, the classifiers' vocabulary did not explicitly capture the error features that were helpful in correctly grading the Korean essay. By overcoming these two limitations, we expect the score range prediction performance to improve.