Earticle

Quality control is difficult to secure objectivity because the quality management of reinforced concrete construction is made by the subjective judgment by experts through a checklist. This study aims to establish an automation process for quality management of reinforced concrete frameworks through a Deep Learning algorithm based on object recognition. Through this, it is possible to save time more objectively than before, and the purpose is to provide intuitive judgment through visualization. This study proposed a quality control process through the learning and verification process with the image data set obtained from AI Hub, and mAP was derived with an accuracy of 0.687. The drone image data of the actual site was determined using the derived algorithm. 3D modeling is performed through the determined drone image to ensure the safety of the inspector and intuitive judgment. The proposed process cannot be confirmed the determined line when matched with a 3D model using PIX4D, but it is judged that it will be applicable to additional processes through the replacement of modeling programs and improvement of Deep-Learning algorithms.

COVID-19 확산으로 인해 헬스장이나 야외보다는 집에서 운동하는 것을 선호하는 사람들이 늘어나고 있다. 홈 트 레이닝을 할 때 잘못된 방법으로 운동을 하게 되면 다칠 수도 있고 운동의 효과가 절감될 수 있는 문제가 있다. 본 논문은 학습에 기반한 포즈 분류 모델을 통해 실시간으로 사용자의 포즈를 분류하고, 사람의 관절 점을 기반으로 포즈를 분석하여 잘못된 포즈를 교정할 수 있도록 도와주는 홈 트레이닝 시스템을 제안한다. 트레이너의 비디오 영상 과 웹캠을 통하여 직접 입력받은 운동 영상을 MediaPipe Pose API를 사용하여 포즈의 관절 좌표를 추출하고, 이 를 데이터 세트로 학습하여 포즈 분류 모델을 구성한다. 사용자는 실시간으로 인식된 본인의 포즈를 확인하면서 운 동할 수 있고 잘못된 포즈에 대해서는 올바른 어드바이스를 제공받을 수 있다. 또한 사용자가 올바른 포즈로 운동을 한 경우 운동 횟수가 증가되어 보다 효율적인 운동이 가능하도록 한다. 맨몸 운동의 대표적 운동인 스쿼트와 푸쉬업 에 대해 본 시스템의 성능을 실험한 결과 스쿼트는 96%이고, 푸쉬업은 95%의 정확도를 보였고 포즈 추론 속도는 평균 31ms가 소요되어 실시간으로 활용 가능함을 알 수 있었다.

Due to the spread of COVID-19, more and more people prefer to exercise at home rather than at gyms or outdoors. If you exercise in the wrong way during home training, you may get hurt and the effectiveness of the exercise may be reduced. This paper proposes a home training system that helps correct wrong posture by classifying user posture in real time through a deep learning-based posture classification model and analyzing posture based on human joint points. The joint coordinates of the pose are extracted using the MediaPipe Pose API for the video images of the trainer and the exercise input images obtained directly from the webcam, and the pose classification model is constructed by learning them as a dataset. The user can exercise while checking his or her pose recognized in real time, and can be provided with the right advice for the wrong pose. In addition, when the user exercises in the right pose, the number of exercises increases, enabling more efficient exercise. As a result of testing the performance of this system for squats and push-ups, which are representative exercises of bare-body exercise, it was found that squats were 96%, push-ups showed 95% accuracy, and pose inference speed took an average of 31 ms, making it available in real-time.

이 연구는 국내 딥러닝 기반 추천 시스템의 최근 연구 문헌을 고찰하기 위해 국내 학술 최대 학술 DB인 학술연구정보서비스(RISS)를 활용하여 체계적 문헌고찰 연구방법을 적용하였다. 문헌 검색을 위해 RISS에 서 ʻ딥러닝ʼ과 ʻ추천 시스템ʼ을 키워드로 추출하고, 이 연구에서 정한 프로토콜(protocol)로 최종 35편의 문헌 을 선정하였다. 선정된 문헌을 바탕으로 국내 딥러닝 기반 추천 시스템의 연구 현황, 연구 주제, 연구 내용, 적용된 딥러닝 모델 등을 분석하여 향후 딥러닝 기반 추천 시스템의 연구 방향을 제시하고자 하였다. 이 연구의 결과로 얻은 결론은 다음과 같다. 첫째, 국내 딥러닝 기반 추천 시스템의 연구가 이미지나 텍스트 등의 정보 분석에서 벗어나 영상과 음성 등을 딥러닝 모델에 적용하는 방법적 연구가 필요하다. 둘째, 추천 시스템의 성능을 향상시키기 위해 보다 다양한 딥러닝 모델을 적용한 연구가 필요하다. 셋째, 기존 연구 경향과 같은 공학적 접근 외에 사회과학적 접근의 연구가 필요하다. 넷째, 교육 분야에서도 딥러닝 기반 추천 시스템에 대한 연구가 필요하다.

In this paper, we propose to review the recent research literature of the recommendation system based on deep learning in Korea, through systematic review method using RISS, the largest academic DB in Korea, To search for studies, we extract 'Deep Learning' and 'Recommendation System' as keyword in the RISS database, select the final 35 studies using the protocol defined in this study, research the current status of deep learning based on recommendation system in Korea, The purpose of this study is to suggest the research direction of domestic deep learning based on recommendation system by analyzing topics, contents, and applied deep learning technology models. The conclusions from this study are as follows. First, the study of recommendation system needs a methodology to apply study to deep learning model by analyzing information such as image and voice out of current image or text. Second, more diverse deep learning models are needed to improve the performance of the recommendation system. Third, the social science approach is needed in addition to the engineering approach as the existing study trend. Fourth, study on deep learning-based recommendation system is also needed in education.

망간은 국내 수요가 많은 광물임에도 불구하고 그동안 가격 예측에 대한 연구는 오랜 기간 이루어지지 않은 상황이다. 그러나 최근 공급망 에 대한 문제가 수면 위로 부상하면서 주요 광물에 대한 공급망 관리를 위한 제도화가 이루어지면서 주요 광물에 대한 국내외 수급동향 및 가격 등에 대한 정보를 요구하고 있다. 본 연구는 이러한 배경하에 핵심광물로 지정된 망간에 대한 가격을 예측하고자 하였다. 2004년 10월부터 2025년 6월까지 망간의 톤당 주별 가격자료를 LSTM 모형에 적용하고 이를 전통적인 가격 예측모형인 ARIMA와 비교하고자 하였다. 분석결과 LSTM 모형의 예측력은 MAPE 2%, MAE 17.80, RMSE 54.72로 비교 방법인 ARIMA에 비해 양호한 예측력을 보였다.

Although manganese is a mineral in high domestic demand, research on price prediction has not been conducted for a long time. However, as supply chain problems have recently emerged to the surface, institutionalization has been made to manage supply chains for major minerals, requiring information on domestic and foreign supply trends and prices for major minerals. Against this background, this study attempted to predict the price of manganese designated as a core mineral. From October 2004 to June 2025, weekly price data per ton of manganese were applied to the LSTM model and compared with ARIMA, a traditional price prediction model. As a result of the analysis, the predictive power of the LSTM model was 2% MAPE, 17.80 MAE, and 54.72 RMSE, showing better predictive power than the comparison method ARIMA.

음성인식 기술은 딥러닝과 결합되며 빠른 속도로 발전하고 있다. 특히 음성인식 서비스가 인공지능 스피커, 차량 용 음성인식, 스마트폰 등의 각종 기기와 연결되며 음성인식 기술이 산업의 특정 분야가 아닌 다양한 곳에 활용되고 있다. 이러한 상황에서 해당 기술에 대한 높은 기대 수준을 맞추기 위한 연구 역시 활발히 진행되고 있다. 그중에서 자연어처리 (NLP, Natural Language Processing)분야에서 음성인식 인식률에 많은 영향을 주는 주변의 소음이나 불필요한 음성신 호를 제거하는 분야에 연구가 필요한 상황이다. 이미 많은 국내외 기업에서 이러한 연구를 위해 최신의 인공지능 기술을 활용하고 있다. 그중에서 합성곱신경망 알고리즘(CNN)을 활용한 연구가 활발하게 진행되고 있다. 본 연구의 목적은 합성 곱 신경망을 통해서 사용자의 발화구간에서 비음성 구간을 판별하는 것으로 5명의 발화자의 음성파일(wav)을 수집하여 학습용 데이터를 생성하고 이를 합성곱신경망을 활용하여 음성 구간과 비음성 구간을 판별하는 분류 모델을 생성하였다. 이후 생성된 모델을 통해 비음성 구간을 탐지하는 실험을 진행한 결과 94%의 정확도를 얻었다.

Speech recognition technology is being combined with deep learning and is developing at a rapid pace. In particular, voice recognition services are connected to various devices such as artificial intelligence speakers, vehicle voice recognition, and smartphones, and voice recognition technology is being used in various places, not in specific areas of the industry. In this situation, research to meet high expectations for the technology is also being actively conducted. Among them, in the field of natural language processing (NLP), there is a need for research in the field of removing ambient noise or unnecessary voice signals that have a great influence on the speech recognition recognition rate. Many domestic and foreign companies are already using the latest AI technology for such research. Among them, research using a convolutional neural network algorithm (CNN) is being actively conducted. The purpose of this study is to determine the non-voice section from the user's speech section through the convolutional neural network. It collects the voice files (wav) of 5 speakers to generate learning data, and utilizes the convolutional neural network to determine the speech section and the non-voice section. A classification model for discriminating speech sections was created. Afterwards, an experiment was conducted to detect the non-speech section through the generated model, and as a result, an accuracy of 94% was obtained.

연구목적: 본 연구는 인공신경망 라이브러리 기술을 이용하여, 기상 데이터 변화 예측을 통한 한반도 가 뭄 취약지역 분석을 목적으로 하였다. 연구방법: 연구지역 중 북한 지역의 다양한 기상데이터의 확보가 힘든 특수성을 고려하여 연구지역의 월별 누적강수량 데이터를 활용하였으며, 통계프로그램 R을 이용 하여 인공신경망 알고리즘을 통한 기상데이터 추정을 수행하였다. 연구결과: 본 논문에서 진행한 연구 결과, 실제 데이터와 예측 데이터 간의 상관계수 값은 인공신경망 알고리즘을 활용한 결과가 회귀분석 결과보다 평균 0.043879 더 높은 것으로 확인되었다. 결론: 연구의 결과는 가뭄 대응을 위한 재난대응 기초 연구 자료로 활용 가능할 것으로 기대한다.

Purpose: In this paper, using artificial neural network algorithm, the Korean Peninsula was analyzed for drought vulnerable areas by predicting weather data changes. Method: Monthly cumulative precipitation data were utilized for research areas considering the specific nature areas, and weather data prediction through artificial neural network algorithm was carried out using statistical program R. The predicted data were applied to the Standardized Precipitation Index (SPI) to analyze drought vulnerable areas in the Korean Peninsula. Result: In this paper, the correlation coefficient values between real and predicted data are found to be 0.043879 higher on average than the regression results, using artificial neural network algorithms. Conclusion: The results of the research are expected to be used as basic research materials for responding to drought.

최근 다양한 분야(Domain)에서 기존의 지식 그래프를 기반으로 해당 분야에 최적화된 새로운 지식 체계를 구축하려는 시도가 활발하게 이루어지고 있다. 하지만 기존의 지식 그래프는 일반적인 개체의 일반적인 관계만을 표현하기 때문에 최근의 팬데믹(pandemic) 상황처럼 새로운 이슈에 대한 지식정보가 빈번하게 추가, 갱신되는 상황에서는 기존의 지식 그래프를 활용한 지식 체계의 정립이 거의 불가능하다는 한계를 갖는다. 이에 본 연구는 이러한 한계를 극복하기 위하여 정형화된 지식 체계를 활용하지 않고 특정 분야의 지식 체계를 구축하는 새로운 방안을 제안한다. 구체적으로는 COVID-19 관점의 개체 사전을 정의하고, BERT를 활용하여 COVID-19 관점의 새로운 사전 학습 모델인 Co-BERT를 구축한다. 이후 이렇게 구축된 Co-BERT 모델의 Self-Attention 가중치를 활용하여 COVID-19 관점의 지식그래프를 생성하는 방법을 제안한다. 본 제안 방법은 특정 도메인에 특화된 지식 그래프를 생성할 수 있는 방법으로, 향후 제안 방법론을 다양한 분야에 적용하여 특화된 지식 체계를 구축하는 연구에 활용할 수 있을 것이라 기대한다.

이 글은 이른바 4차 산업혁명 시대를 맞이하여 현 단계 인공지능(AI)이 쓴 문학작품의 문제를 인간성(humanity) 구현의 측면에서 파악하고, 나아가 인공지능의 문학창작의 미래를 전망해 보기 위한 것이다. 이를 위한 논의는 인공지능의 인간성 구현에 초점을 두되, 인간성의세 가지 속성인 3ty, 즉 창의성(creativity), 감성(affectivity/sensibility), 윤리성(morality)의측면에서 인공지능이 쓴 문학작품들을 검토하고자 했다. 인공지능이 쓴 소설과 시나리오는 아직은 인간이 창작한 소설에 비해 상당한 한계를 가지고 있다. 인공지능이 빅데이터에 의한 자료 입력, 딥러닝 기반 언어처리와 논리적 추론을 통해소설과 시나리오 쓰기는 가능한 단계에 이르렀지만, 인간의 개입 없이 100% 독자적인 소설창작은 불가능하다. 기존 소설과 시나리오의 서사 패턴을 분석하여 논리적 추론을 함으로써일정한 분량의 글을 쓸 수 있지만, 그 분량은 매우 짧고 아직 ‘창의적 모방’ 수준을 넘어서지못했다. 따라서 문학 창작을 하는 인공지능을 창작의 주체로 인정할 수 있는 단계에 있지 않다. 인공지능의 시 창작도 소설과 시나리오 창작의 경우와 크게 다르지 않았다. 인공지능의시 창작은 이외로 소설과 시나리오 창작의 경우보다 앞서 이루어졌다. 특정 시인의 시 작품을딥러닝으로 학습한 인공지능이 제시된 사진을 보고 일정한 분량의 시를 창작하거나 인간이제시한 문장에 반응하는 형식으로 시를 창작하는 단계에 있다. 그렇지만 역시 인간의 도움 없이 주체적인 시 창작이 이루어지는 단계는 아니다. 미래 인공지능에 의한 문학 창작은 인간과 같은 창의성을 구현하는 것은 물론 인간적인감성과 윤리성을 토대로 신속하고 다양한 문학 창작이 가능한 수준으로 발전할 수 있다. 그런데 인공지능의 창의성 구현은 비인간적이고 반윤리적인 방향으로 나아갈 위험도 있다. 이에대한 경계심과 적절한 규제가 필요하겠지만, 이제부터 인간과 기계의 협력과 상호 소통을 통한 공생을 추구하는 탈인간주의(Post-Humanism)의 철학을 가져야 할 때이다.

This article aims to understand problems of literary works written by AI at the present stage of the 4th Industrial Revolution in terms of the implementation of humanity and to examine the future of AI's creation. This discussion focuses on the humanity of AI. For this I examined literary works written by AI in terms of three attributes of humanity: creativity, affectivity/sensibility, and morality. Short stories and scenarios written by AI still have considerable limitations compared with man-made works. Although AI has been able to write short stories and scenarios through data input, Deep-Learning based language processing and logical reasoning, it is impossible to create 100% original work without human intervention. It is possible to write a certain amount of articles by analyzing the narrative patterns of existing works and making logical reasoning, but the amount is very short and has not yet exceeded the level of ‘creative imitation’. Therefore, it is not at a stage to recognize AI as the subject of creation. The poetry has a strong subjective nature and emotional action among artistic branches of literature. The poetry creation of AI was done in two ways. One way is that AI creates a poem by seeing photographs presented by human after learned poems of a poet by Deep-Learning. Another way is that AI completes a piece of poem in a way which responds to a line of poem which a person presents. However, it is not a stage in which subjective creation of poetry is done without human help. The AI-based literary creation will develop rapidly in the near future. It is possible to develop creativity that is human-like as well as rapid and diverse literary creation based on human sensitivity and morality. A professional AI artist will appear in the near future. However, the creativity of AI is also in danger of being inhuman and anti-ethical. Appropriate regulations and discipline are naturally necessary. From now on, it is necessary for us to overcome the humanism of rejecting the machine or perceiving the machine as a means and a tool in the human-centered viewpoint. We need to have the post-humanism in pursuit of symbiosis through cooperation and mutual communication between man and machine.

Free space optical (FSO) communication systems offer high-bandwidth, secure data transmission over wireless channels. Recent advancements in machine learning (ML) and deep learning (DL) have considerable promise in mitigating these challenges and enhancing the reliability and efficiency of FSO systems. This comprehensive survey examines ML and DL techniques applied to FSO systems, covering advancements in channel modeling and estimation, and demodulation. Additionally, this review highlights the role of ML and DL in hybrid FSO/RF systems, focusing on resource management, dynamic switching, relay selection, underwater FSO, and ATP. Emerging trends, future research directions, standardization efforts, and unresolved challenges are discussed. Our overall conclusion highlights that DL, especially hybrid and attention-based models, demonstrates strong potential in dynamic channel adaptation and tracking under turbulence, while reinforcement learning shows promise for real-time resource allocation and switching.

Sepsis and Neonatal sepsis are major challenges in global healthcare because they cause life-threatening organ dysfunction in intensive care adult and pediatric patients due to downregulated host response to a particular infection. Early clinical identification of sepsis is difficult, and failure to provide prompt treatment can often lead to crucial stages and increase the rates of fatality. Thus an intense study is needed to determine and categorize sepsis in its initial stage. The complexity of varying clinical statistics makes it difficult to attain a precise definition in pediatrics. The advanced Machine Learning (ML) and Deep Learning (DL) technologies in the implementation of protocols show promising real-time models for predicting sepsis at the primary stage and thereby reducing the mortality rate. This review article contemplates the complete list of procedures through which sepsis and neonatal sepsis are speculated by ML and DL and concentrates specifically on data available in the adult emergency care unit as well as the neonatal intensive care unit. The survey process was carried out by searching terms related to ML and DL merged with topics concerning sepsis and neonatal sepsis. The literature analysis was carried out from Scopus, Web of Science, and PubMed databases for the period from 2015 to 2022. The assessment of the risk of bias was carried out for the eleven selected papers using the Prediction Model Risk of Bias Assessment Tool (PROBAST). The eleven papers were selected from different medical care units based on the performance measure AUROC, which ranges from 0.68 to 0.95. Five papers involving ML/DL models reduce the bias and lessen risk occurrence. Five papers generate an increase in bias but can be applied to new data. One paper works with above twenty-five features has high-risk probability but predicts patients within 5?6 h in the future. This survey portrays the role of prediction models that supports the researchers and clinicians for better decision-making and antibiotic administration at an earlier stage.

There is a great demand for an efficient security framework which can secure IoT systems from potential adversarial attacks. However, it is challenging to design a suitable security model for IoT considering the dynamic and distributed nature of IoT. This motivates the researchers to focus more on investigating the role of machine learning (ML) in the designing of security models. A brief analysis of different ML algorithms for IoT security is discussed along with the advantages and limitations of ML algorithms. Existing studies state that ML algorithms suffer from the problem of high computational overhead and risk of privacy leakage. In this context, this review focuses on the implementation of federated learning (FL) and deep learning (DL) algorithms for IoT security. Unlike conventional ML techniques, FL models can maintain the privacy of data while sharing information with other systems. The study suggests that FL can overcome the drawbacks of conventional ML techniques in terms of maintaining the privacy of data while sharing information with other systems. The study discusses different models, overview, comparisons, and summarization of FL and DL-based techniques for IoT security.

Cardio-Vascular Diseases (CVD) is found to be rampant in the populace leading to fatal death. The statistics of a recent survey reports that the mortality rate is expanding due to obesity, cholesterol, high blood pressure and usage of tobacco among the people. The severity of the disease is piling up due to the above factors. Studying about the variations of these factors and their impact on CVD is the demand of the hour. This necessitates the usage of modern techniques to identify the disease at its outset and to aid a markdown in the mortality rate. Artificial Intelligence and Data Mining domains have a research scope with their enormous techniques that would aassist in the prediction of the CVD priory and identify their behavioural patterns in the large volume of data. The results of these predictions will help the clinicians in decision making and early diagnosis, which would reduce the risk of patients becoming fatal. This paper compares and reports the various Classification, Data Mining, Machine Learning, Deep Learning models that are used for prediction of the Cardio-Vascular diseases. The survey is organized as threefold: Classification and Data Mining Techniques for CVD, Machine Learning Models for CVD and Deep Learning Models for CVD prediction. The performance metrics used for reporting the accuracy, the dataset used for prediction and classification, and the tools used for each category of these techniques are also compiled and reported in this survey.

Reliable channel estimation is critical for achieving high-speed and energy-efficient communication in Unmanned Aerial Vehicle-Free Space Optical (UAV-FSO) systems, particularly under dynamic impairments such as atmospheric turbulence (AT) and pointing errors (PEs). This paper proposes a pilot-free channel estimation framework based on a Variational Autoencoder (VAE). The system employs Intensity Modulation/Direct Detection (IM/DD) with?-ary one-hot encoded symbols (). The VAE encodes noisy received signals into a 128-dimensional latent space and reconstructs the transmitted data, effectively learning the joint effects of AT, PEs, and AWGN. Unlike prior works that primarily consider boresight or Gaussian-jitter PEs, this study explicitly incorporates a Nakagami-modeled PE distribution, capturing UAV-induced beam misalignment under mobility, vibration, and turbulence coupling. Simulation results show that the proposed VAE significantly outperforms conventional estimators (LS, MMSE, LMMSE) and deep learning baselines (AE, DNN, CNN) across various turbulence strengths. Under strong turbulence and PEs, the VAE attains nearly two-fold lower MSE compared to CNN and DNN. In addition, evaluation on real turbulence-impaired datasets further validates robustness and generalization. The proposed pilot-free scheme delivers accurate channel estimation, reduced BER, and improved spectral efficiency, making it suitable for real-time adaptive UAV-FSO communication.

Diabetic Retinopathy (DR) is a leading cause of vision impairment and blindness worldwide. Early diagnosis is crucial for preventing irreversible vision loss, but manual screening methods are time-consuming and often inconsistent. Deep learning (DL) techniques have shown promise in automating DR detection; however, many existing models still struggle to capture subtle lesions and distinguish fine-grained severity stages. In this survey, we comprehensively review recent DL-based approaches for DR classification, emphasizing attention mechanisms, feature fusion strategies, and stage-wise grading. To address current gaps, we propose a hybrid taxonomy that identifies effective combinations such as texture-based attention, CNN-Transformer fusion, and multi-modal integration. Additionally, we validate our previously published model, STMFNet, a spatial texture-aware attention network based on EfficientNet, across four benchmark datasets. On EyePACS and Messidor, STMFNet achieves up to 98.10% accuracy, outperforming several state-of-the-art (SOTA) models under similar settings. This study provides both a consolidated overview of DR detection advancements and a practical benchmark framework to guide future research in AI-assisted DR classification.