HJHyeonseok Jin

Hyeonseok Jin

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hyeonseok Jin has a total of 6 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    An Improved Small Defect Classification System Using Image Patching Methodology on Aluminum Car Doors

    by Ugur Ercelik, Hyeonseok Jin, Longfei Li, Kyungbaek Kim

    Small surface defects in automotive manufacturing such as micro-cracks, springbacks, and minor deformations typically appear in highly curved or reflective regions of large and complex components. Although recent Artificial Intelligence (AI)–based visual inspection systems have improved classification, existing whole-image classification methods still struggle to identify subtle or localized defects that occupy only a small portion of the image. These subtle defects occupy only small and localized regions, analyzing the entire image as a single representation often leads to missed detections and insufficient sensitivity. To address these issues, we propose a small defect classification system using patch-based methodology. The approach divides each car-door image into spatial windows, enabling localized inspection of subtle defects that would otherwise be masked in full-image representations. First, Each window is processed using an enhanced Large Kernel Attention (LKA)-based ResNet50 backbone, where the LKA mechanism increases the effective receptive field and improves small-defect discriminability within each patch. Second, To overcome the limitation of independent patch predictions, we introduce a Self-Attention-Based Multi-Layer Perceptron (MLP) Aggregator that learns inter-patch dependencies and adaptively fuses window-level outputs into a robust whole-image decision, ensuring that cross-window defects are not overlooked. Experimental results showthat the combination of localized patch inspection, LKA-enhanced feature extraction, and self-attention based MLP aggregation yields superior performance, achieving 95% accuracy and 95% F1-score. These results confirm the effectiveness of the proposed image patching methodology as a reliable and scalable solution for industrial small-defect detection.


  • Open Access

    A Data-Driven Model to Predict Regular Strength Exercise Patterns in Patients With Chronic Obstructive Pulmonary Disease: A Secondary Analysis

    by Hyeonseok Jin, Ja Yun Choi, Sehyun Cho, Kyungbaek Kim

    Regular exercise plays a crucial role in the management of chronic obstructive pulmonary disease (COPD), yet many patients struggle with adherence to such health-promoting behaviors. This study aimed to develop and evaluate machine learning and deep learning models to predict adherence to regular strength exercise among individuals with COPD using data from the Korean National Health and Nutrition Examination Survey (KNHANES) collected between 2007 and 2019. A total of 5,060 patients with COPD were identified based on established clinical criteria, and 25 key variables were selected for model development. We compared the performance of four machine learning models (logistic regression, support vector machine, decision tree, XGBoost) and three deep learning models (multi-layer perceptron, convolutional neural network, FT-Transformer). Model performance was assessed using precision, recall, F1 score, and weighted F1 score. The FT-Transformer model outperformed other models, achieving a precision of 0.64, a recall of 0.68, an F1 score of 0.64, and a weighted F1 score of 0.71. Shapley Additive Explanations (SHAP) were used to interpret model predictions and revealed that weekly walking frequency, education level, and weight management were key predictors of adherence. These findings demonstrate the potential of AI-driven models to provide actionable insights into patient behavior and support personalized interventions. Further research using longitudinal data is recommended to improve predictive performance and validate model applicability in clinical settings.


  • Open Access

    DATC-STP: Towards Accurate yet Efficient Spatiotemporal Prediction With Transformer-Style CNN

    by Hyeonseok Jin, Kyungbaek Kim

    Recently, convolutional neural networks (CNNs) or vision transformers (ViTs) based Multi-In-Multi-Out (MIMO) architectures are proposed to overcome the limitations of recurrent neural networks (RNNs) based Single-In-Single-Out (SISO) architectures. These architectures prevent the inherent limitations of RNNs, which degrade performance and inefficiency of parallelization due to the sequential properties. However, there are still some challenges. CNN-based MIMO architectures have difficulty capturing global spatiotemporal information due to the local properties of its kernel. Meanwhile, ViT-based MIMO architectures have difficulty capturing local spatiotemporal information and require high-computational resource due to the self-attention. To improve MIMO architecture with overcome these limitations, we propose a novel accurate yet efficient Dual-Attention Transformer-style CNN for Spatiotemporal Prediction (DATC-STP).DATC-STP captures both local and global spatiotemporal information by 3D patch embedding and Transformer-style CNN. Specifically, 3D patch embedding extract local spatiotemporal features and reduce the size of input data including temporal, height, and width. Two Transformer-style CNN based attention blocks treat spatiotemporal data similarly with image and capture global information with CNNs. These structure makes DATC-STP accurate yet efficient. To demonstrate the effectiveness of DATC-STP, we conduct comprehensive experiments with three promising benchmark datasets, MovingMNIST, TaxiBJ, and KTH.We evaluated that the proposed DATC-STP achieves both competitive performance and efficient. Furthermore, results of ablation study demonstrates the useful for each component of DATC-STP and highlights the potential of proposed methods.


  • Open Access

    Enhancing deep learning models for predicting smoking Status using clinical data in patients with chronic obstructive pulmonary disease

    by Sehyun Cho, Hyeonseok Jin, Kyungbaek Kim, Sola Cho, Ja Yun Choi

    Objective: This study aimed to develop and evaluate deep learning models to improve the prediction of persistent smoking in patients with chronic obstructive pulmonary disease (COPD) by integrating behavioral and psychosocial variables with clinical data from a structured national dataset. Methods: Three deep learning models and one machine learning model were developed and assessed using clinical, behavioral, and psychosocial data from 350 patients with COPD, including 51 current smokers. Data preprocessing involved imputation, variable transformation, and class weighting. Hyperparameter optimization was performed using the Optuna framework. Model performance was evaluated with repeated stratified K-fold cross-validation, and the macro F1 score was the primary metric. Shapley Additive Explanations (SHAPs) were applied to assess feature importance and improve interpretability. Results: The Residual Neural Network achieved the highest performance, with a macro F1 score of .87 (95% confidence interval: .83–.89). SHAP analysis highlighted professional advice to quit, employment status, sputum symptoms lasting more than 3 months, perceived stress level, health check-up experience, and health literacy as key predictors of persistent smoking. Conclusion: Incorporating behavioral and psychosocial data enabled the models to capture complex smoking patterns while maintaining interpretability. These findings emphasize the value of multidimensional data in identifying high-risk individuals and informing targeted smoking cessation strategies in COPD care. Future research should include synthesized behavioral variables often absent from large external datasets and validate model performance in more diverse populations.