Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jin-Woo Jung has a total of 65 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Warehouse Mobile Robot Path Planning Performance Sensitivity to the Neighbor Radius Parameter

    by Jihong Jeong, Jin-Woo Jung

    Many RRT*-based sampling path planning algorithms consider neighboring nodes around a newly added node. The neighbor radius parameter rnear determines which nodes are included. The performance of RRT*-based algorithms can vary significantly with rnear. This variation can weaken generalization across environments. This paper quantitatively analyzes the effect of rnear on performance in sampling-based path planning for mobile robots in a warehouse environment. We evaluate RRT*-based algorithms by varying rnear. We then select the heuristic chosen rnear for each algorithm and compare the algorithms under the same conditions. Experiments are conducted in a warehouse environment with a fixed start position and five goal positions. Performance is evaluated using planning time, path length, and cumulative change in turning angle. Lower values indicate better performance for all three metrics. Based on the experimental results, we derive a heuristic value of rnear for each case. We also identify algorithm characteristics in computational efficiency and path quality under the heuristically chosen parameter settings. The final goal of this study is to provide quantitative evidence for selecting rnear in warehouse applications. We also present guidelines for parameter setting and algorithm selection for RRT*-based sampling path planning.


  • Open Access

    Structure-Aware Pixel Art Scaling via Block Size Detection

    by Jun Won Seo, Jun Won Lee, Jong Hyuck Lee, Jun Beom Kim, Jin-Woo Jung

    Standard interpolation methods degrade pixel art through blurring or geometric distortion. We propose a lossless scaling algorithm that detects the intrinsic block size to normalize the image grid, thereby expanding the set of valid scaling factors beyond standard integer multiples. This approach enables precise, distortion-free resizing closer to user-specified scales. To validate this approach, we introduce a novel evaluation framework consisting of Color Loss (CL), Block Size Consistency (BSC), and reversibility (REV) tests. Experimental results demonstrate that the proposed method maintains the original palette and grid structure without introducing interpolation artifacts. Furthermore, the reversibility tests confirm that the scaling process remains mathematically lossless, ensuring the genre’s structural and chromatic integrity.

  • Open Access

    Feasibility of Drone-Mounted Camera for Real-Time MA-rPPG in Smart Mirror Systems

    by Mohammad Afif Kasno, Yong-Sik Choi, Jin-Woo Jung

    Remote photoplethysmography (rPPG) enables contactless estimation of cardiovascular signals from video, but most existing studies assume a fixed, stationary camera. This study investigates the feasibility of performing real-time moving-average rPPG (MA-rPPG) using a drone-mounted camera, where platform motion, vibration, and viewing distance introduce additional challenges. Building on our previously validated real-time MA-rPPG smart mirror platform, we reuse the smart mirror interface as a unified frontend for visualization, synchronization, and logging while adapting the MA-rPPG pipeline to operate on live video streamed from an off-the-shelf DJI Tello micro-drone. Feasibility experiments were conducted with 10 participants under controlled indoor lighting and constrained flight conditions, where the drone maintained a stable hover in front of a standing subject and facial video was processed in real time to estimate heart rate from a forehead region of interest. To avoid cross-modality bias and clarify the effect of the aerial imaging platform, drone-derived MA-rPPG outputs were compared against a fixed desktop-camera MA-rPPG reference using the same trained model, enabling a controlled, like-for-like evaluation. The results indicate that continuous heart-rate estimation from a drone camera is feasible in our controlled hover-only setup, while agreement tended to vary with hover stability and effective facial resolution. This work is presented strictly as a feasibility-stage investigation and does not claim clinical validity. The findings provide an experimental baseline and operating-envelope insight for future motion-robust rPPG on mobile and aerial health-sensing platforms.


  • Open Access

    Exploring a New Architecture for Efficient Parameter Fine-Tuning in SLoRA Multitasking Scenarios

    by Ce Shi, Jin-Woo Jung

    Propose an enhanced LoRA (Low-Rank Adaptation) MoE (mixed expert) architecture, SLoRA (Enhanced LoRA MoE Architecture), aimed at addressing the key problem of efficient parameter fine-tuning in multitasking scenarios. Given the high cost of traditional full fine-tuning as the parameter size of visual language models increases, and the limitations of LoRA as a popular PEFT (parameter-efficient fine-tuning) method in multitasking, such as inadequate adaptability and difficulty in capturing complex task patterns, as well as the catastrophic forgetting and knowledge fragmentation challenges faced by existing research on integrating mixed expert (MoE) mechanisms into LoRA, SLoRA utilizes orthogonal constraint optimization to reduce disturbance to existing knowledge through constraint solution space initialization, alleviating catastrophic forgetting (old task accuracy retention rate reaches 92.4%, 16.1% higher than LoRA), and an optimized MoE structure that includes general experts (retaining pre-trained knowledge) and task-specific experts (dynamic routing adaptation tasks) to enhance multitask adaptability. Experimental results show that in commonsense reasoning tasks, SLoRA’s accuracy is 9.0% higher than LoRA and 3.7% higher than AdaLoRA on the WSC dataset, and its F1 score is 7.7% higher than LoRA and 2.9% higher than AdaLoRA on the CommonsenseQA dataset; in multimodal tasks, its average score is up to 15.3% higher than LoRA, demonstrating significant advantages over existing methods.

  • Open Access

    Fast Path-Planning Algorithm for Mobile Robots via Straight Strategy

    by Jihong Jeong, Jin-Woo Jung

    RRT*, which augments the RRT with the ChooseParent and Rewire steps, is a widely used algorithm for path planning. RRT*-based algorithms are effective for improving the quality of paths in the Rewire step. However, as the expansion continues, there are more nodes to be inspected, which can slow down the path search. In addition, due to the parameters set in the rewire step, a trade-off issue between path quality and planning time arises. In this paper, we propose Straight-RRT to improve upon these limitations. Straight-RRT applies the Straight strategy to rapidly obtain an initial path and then returns a refined path using the MoveParent strategies. Accordingly, Straight-RRT adopts the following mechanisms. (1) The Straight strategy is applied for rapidly finding an initial path. This procedure quickly finds a feasible initial path. (2) MoveParent is a strategy inspired by parametric equations for path optimization. This complements the weaknesses of the Straight strategy and the limitations of the triangle inequality. The MoveParent strategy is applied bidirectionally. These procedures progressively refine the path and improve efficiency. We propose an algorithm that is faster than other algorithms using these strategies and minimizes the trade-off caused by parameter settings. In the experimental comparison results across most environments, our approach achieved shorter planning times than the compared baseline algorithms and produced paths of comparable quality.


  • Open Access

    Deep Learning-Based Human Activity Recognition Using Dilated CNN and LSTM on Video Sequences of Various Actions Dataset

    by Bakht Alam Khan, Jin-Woo Jung

    Human Activity Recognition (HAR) plays a critical role across various fields, including surveillance, healthcare, and robotics, by enabling systems to interpret and respond to human behaviors. In this research, we present an innovative method for HAR that leverages the strengths of Dilated Convolutional Neural Networks (CNNs) integrated with Long Short-Term Memory (LSTM) networks. The proposed architecture achieves an impressive accuracy of 94.9%, surpassing the conventional CNN-LSTM approach, which achieves 93.7% accuracy on the challenging UCF 50 dataset. The use of dilated CNNs significantly enhances the model’s ability to capture extensive spatial–temporal features by expanding the receptive field, thus enabling the recognition of intricate human activities. This approach effectively preserves fine-grained details without increasing computational costs. The inclusion of LSTM layers further strengthens the model’s performance by capturing temporal dependencies, allowing for a deeper understanding of action sequences over time. To validate the robustness of our model, we assessed its generalization capabilities on an unseen YouTube video, demonstrating its adaptability to real-world applications. The superior performance and flexibility of our approach suggests its potential to advance HAR applications in areas like surveillance, human–computer interaction, and healthcare monitoring.


  • Open Access

    Feasibility of an AI-Enabled Smart Mirror Integrating MA-rPPG, Facial Affect, and Conversational Guidance in Realtime

    by Mohammad Afif Kasno, Jin-Woo Jung

    This paper presents a real-time smart mirror system combining multiple AI modules for multimodal health monitoring. The proposed platform integrates three core components: facial expression analysis, remote photoplethysmography (rPPG), and conversational AI. A key innovation lies in transforming the Moving Average rPPG (MA-rPPG) model—originally developed for offline batch processing—into a real-time, continuously streaming setup, enabling seamless heart rate and peripheral oxygen saturation (SpO2) monitoring using standard webcams. The system also incorporates the DeepFace facial analysis library for live emotion, age detection, and a Generative Pre-trained Transformer 4o (GPT-4o)-based mental health chatbot with bilingual (English/Korean) support and voice synthesis. Embedded into a touchscreen mirror with Graphical User Interface (GUI), this solution delivers ambient, low-interruption interaction and real-time user feedback. By unifying these AI modules within an interactive smart mirror, our findings demonstrate the feasibility of integrating multimodal sensing (rPPG, affect detection) and conversational AI into a real-time smart mirror platform. This system is presented as a feasibility-stage prototype to promote real-time health awareness and empathetic feedback. The physiological validation was limited to a single subject, and the user evaluation constituted only a small formative assessment; therefore, results should be interpreted strictly as preliminary feasibility evidence. The system is not intended to provide clinical diagnosis or generalizable accuracy at this stage.


  • Open Access

    Enhanced Graph Autoencoder for Graph Anomaly Detection Using Subgraph Information

    by Chi Zhang, Jin-Woo Jung

    Graph anomaly detection aims at identifying rare, unusual entities in attributed networks with respect to their patterns or structures that deviate significantly from the majority within a graph. Over the years, extensive efforts in this field have been dedicated to the powerful capability of attributed networks to model real-world systems. Given the scarcity of labeled anomalies, current research primarily emphasizes model design via unsupervised learning. Graph autoencoders have been widely utilized for such purposes, leveraging the outstanding capabilities of Graph Neural Networks to model graph structured data. However, most existing graph autoencoder-based anomaly detectors do not exploit the nodes’ local subgraph information, limiting their ability to comprehensively understand the network for better representation learning. Moreover, these methods place greater emphasis on the attribute reconstruction process while neglecting the structure reconstruction aspect. This paper proposes an enhanced graph autoencoder framework for graph anomaly detection tasks that incorporates a subgraph extraction and aggregation preprocessing stage to utilize the nodes’ local topological information for enhanced embedding generation and to induce an additional node–subgraph view through model learning. A graph structure learning-based decoder is introduced as the structure decoder for better relationship learning. Finally, during the anomaly scoring stage, a node neighborhood selection technique is applied to enhance the detection performance. The effectiveness of the proposed framework is demonstrated through comprehensive experiments conducted on six commonly used real-world datasets.


  • Open Access

    Deep Learning-Based Human Activity Recognition Using Dilated CNN and LSTM on Video Sequences of Various Actions Dataset

    by Bakht Alam Khan, Jin-Woo Jung

    Human Activity Recognition (HAR) plays a critical role across various fields, including surveillance, healthcare, and robotics, by enabling systems to interpret and respond to human behaviors. In this research, we present an innovative method for HAR that leverages the strengths of Dilated Convolutional Neural Networks (CNNs) integrated with Long Short-Term Memory (LSTM) networks. The proposed architecture achieves an impressive accuracy of 94.9%, surpassing the conventional CNN-LSTM approach, which achieves 93.7% accuracy on the challenging UCF 50 dataset. The use of dilated CNNs significantly enhances the model's ability to capture extensive spatial-temporal features by expanding the receptive field, thus enabling the recognition of intricate human activities. This approach effectively preserves fine-grained details without increasing computational costs. The inclusion of LSTM layers further strengthens the model's performance by capturing temporal dependencies, allowing for a deeper understanding of action sequences over time. To validate the robustness of our model, we assessed its generalization capabilities on an unseen YouTube video, demonstrating its adaptability to real-world applications. The superior performance and flexibility of our approach suggests its potential to advance HAR applications in areas like surveillance, human-computer interaction, and healthcare monitoring.

  • Open Access

    Affordable 3D Orientation Visualization Solution for Working Class Remotely Operated Vehicles (ROV)

    by Mohammad Afif Kasno, Izzat Nadzmi Yahaya, Jin-Woo Jung

    ROV operators often encounter challenges with orientation awareness while operating underwater, primarily due to relying solely on 2D camera feeds to manually control the ROV robot arm. This limitation in underwater visibility and orientation awareness, as observed among Malaysian ROV operators, can compromise the accuracy of arm placement, and pose a risk of tool damage if not handle with care. To address this, a 3D orientation monitoring system for ROVs has been developed, leveraging measurement sensors with nine degrees of freedom (DOF). These sensors capture crucial parameters such as roll, pitch, yaw, and heading, providing real-time data on the ROV’s position along the X, Y, and Z axes to ensure precise orientation. These data are then utilized to generate and process 3D imaging and develop a corresponding 3D model of the operational ROV underwater, accurately reflecting its orientation in a visual representation by using an open-source platform. Due to constraints set by an agreement with the working class ROV operators, only short-term tests (up to 1 min) could be performed at the dockyard. A video demonstration of a working class ROV replica moving and reflecting in a 3D simulation in real-time was also presented. Despite these limitations, our findings demonstrate the feasibility and potential of a cost-effective 3D orientation visualization system for working class ROVs. With mean absolute error (MAE) error less than 2%, the results align with the performance expectations of the actual working ROV.