CPChulwoo Pack

Chulwoo Pack

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Chulwoo Pack has a total of 25 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Trustworthy Multimodal Learning via Uncertainty-aware Feature Modulation and Reliability Auditing

    by Seungpil Choi, Mukhtiar Ali, Chulwoo Pack, Eungjoo Lee

    Wearable physiological monitoring is increasingly used for stress and affect recognition (e.g., driver-state and daily-life stress tracking), yet fusing heterogeneous sensor streams remains fragile when individual channels degrade or fail without warning. We present a multimodal physiological fusion framework that pairs uncertainty-aware feature modulation with a post-hoc trustworthiness auditing mechanism. Modality-specific Transformer encoders are first aligned through auxiliary classification and mutual distillation so that the resulting representations support stable uncertainty estimation. Noise-perturbation sensitivity scores then down-weight unreliable modalities before a multimodal Transformer produces the fused prediction, with cyclic optimization and an EMA parameter update regularizing the training dynamics. At inference, a trustworthiness auditing scheme measures epistemic honesty under progressive sensor corruption and quantifies modality reliance under channel dropout, providing a sensor-level diagnostic of when and why the model's outputs should be trusted. Across WESAD and AffectiveROAD, our method achieves accuracy and F1 competitive with the compared methods while additionally auditing its own uncertainty behavior; on WESAD, where sensor corruption is resolvable, it also distributes its reliance more evenly across sensors and responds more honestly to input degradation.

  • Open Access

    Real-Time Vehicle Speed Gun With LiDAR–Camera Fusion

    by Liem H. T. Nguyen, Moditha Reddy Jagannatha Reddy, Chulwoo Pack, Sungyong Jung, Hyusim Park

    This paper presents a dual-sensor speed measurement system that fuses LiDAR-based distance ranging with depth-camera tracking through a Kalman filter. We quantitatively evaluate the separate contributions of per-modality filtering and dual-modal fusion via outdoor vehicle experiments encompassing acceleration, deceleration, and steady-state driving conditions. Per-sensor filtering reduces distance error and suppresses high-frequency measurement jitter, while sensor fusion further lowers the root-mean-square error (RMSE) and produces speed profiles that closely match ground-truth measurements. The proposed system operates in real time on a Raspberry Pi platform. Time-stamped CSV speed logs and synchronized, annotated video from real vehicle trials are provided to enable independent validation. The results demonstrate that complementary sensing combined with probabilistic sensor fusion offers a practical and deployable solution for accurate speed measurement under real-world conditions. The fused system achieves a speed root-mean-square error (RMSE) of 0.4993 mph and a distance RMSE of 0.4339 m, representing improvements of 55% and 56%, respectively, over the raw single-sensor baselines. The primary novelty of this research is the development of a three-stage processing architecture that systematically isolates and quantifies these performance gains achieved by per-modality Kalman filtering and subsequent probabilistic sensor fusion. Furthermore, this architecture is implemented and empirically validated on a resource-constrained Raspberry Pi platform under dynamic, real-world driving conditions.


  • Open Access

    Deep learning-based classification of nitrate and nitrite concentrations from water samples using colorimetric test strip images

    by Muhammad Roman, Mazhar Sher, Chamika Kuruppuarachchi, Arshid Ali, Chulwoo Pack, Azlan Zahid, Ali Mirzakhani Nafchi

    Accurate monitoring of nitrate and nitrite concentrations in water is essential for sustainable agriculture, safeguarding public health, and protecting aquatic ecosystems from nutrient pollution. Traditional methods for detecting nitrate and nitrite in water samples are precise but costly, complex, and time-consuming, limiting their practicality for frequent on-site testing. This research proposes deep learning-based computer vision techniques to classify nitrate and nitrite concentrations using images of colorimetric test strips. An RGB IMX219 camera was used to acquire images of colorimetric test strips under standardized, controlled illumination conditions to ensure consistent image quality. A total of 1938 nitrate images and 1190 nitrite images were collected before augmentation. After preprocessing and training-only data augmentation, both classical machine learning baselines based on hand-crafted color and texture features and deep learning models—including a multilayer perceptron (MLP) and convolutional neural networks (AlexNet, VGG16, ResNet18, and GoogLeNet)—were trained and evaluated using an independent test set and stratified fivefold cross-validation. For nitrate classification, ResNet18 and GoogLeNet achieved near-perfect 100% test accuracy, with mean cross-validation accuracy of 99.97% ± 0.04%, substantially outperforming classical baseline models based on hand-crafted color and texture features, which achieved at most 83.5% test accuracy. For nitrite classification, GoogLeNet achieved the strongest overall performance, with a test accuracy of 97.48% and a fivefold cross-validation accuracy of 95.22% ± 1.17%, substantially outperforming the best classical baseline model, which achieved a maximum test accuracy of 83.19%. These results demonstrate that deep CNN-based feature learning provides a significant performance advantage over simpler methods under controlled imaging conditions, supporting the suitability of the proposed system for rapid, image-based water quality assessment and motivating future evaluation under broader real-world deployment scenarios.


  • Open Access

    MM-SCORE: A Two-Factor Framework for Auditing Multimodal Dataset Quality

    by Omeshamisu Anigala, Joy Nnenna Okolo, Junggab Son, Chulwoo Pack

    Multimodal learning has emerged as a transformative approach in applying artificial intelligence (AI) to various industries, such as construction and manufacturing, because it enables the integration of diverse data sources, including video surveillance, sensor readings, audio signals, and textual reports, into unified predictive and monitoring models. Multimodal models are only as reliable as the datasets behind them. We present MM-SCORE, a framework for auditing and improving multimodal datasets (MMD) along two factors: noise and alignment. MM-SCORE turns these factors into actionable scorecards and a targeted repair loop that prioritizes edits by their measured impact on downstream performance. Across image-text and video corpora, it surfaces hidden pathologies—prompt-caption drift, temporal mismatch—and its guided fixes consistently lift accuracy and robustness without changing model architectures. The factors match cleanly with the regulatory notions of accuracy, representativeness, completeness, and relevance, making the quality of the dataset auditable, comparable, and ultimately fixable.


  • Open Access

    A Novel Dataset for Flood Detection Robust to Seasonal Changes in Satellite Imagery

    by Youngsun Jang, Dongyoun Kim, Chulwoo Pack, Kwanghee Won

    This study introduces a novel dataset for segmenting flooded areas in satellite images. After reviewing 77 existing benchmarks utilizing satellite imagery, we identified a shortage of suitable datasets for this specific task. To fill this gap, we collected satellite imagery of the 2019 Midwestern USA floods from Planet Explorer by Planet Labs (Image \c{opyright} 2024 Planet Labs PBC). The dataset consists of 10 satellite images per location, each containing both flooded and non-flooded areas. We selected ten locations from each of the five states: Iowa, Kansas, Montana, Nebraska, and South Dakota. The dataset ensures uniform resolution and resizing during data processing. For evaluating semantic segmentation performance, we tested state-of-the-art models in computer vision and remote sensing on our dataset. Additionally, we conducted an ablation study varying window sizes to capture temporal characteristics. Overall, the models demonstrated modest results, suggesting a requirement for future multimodal and temporal learning strategies. The dataset will be publicly available on .

  • Open Access

    Leveraging Textual Memory and Key Frame Reasoning for Full Video Understanding Using Off-the-Shelf LLMs and VLMs (Student Abstract)

    by Harsh Dubey, Chulwoo Pack

    To address the limitations of current Large-scale Video-Language Models (LVLMs) in fine-grained understanding and long-term temporal memory, we propose a novel video understanding approach that integrates a Vision Language Model (VLM) and a Large Language Model (LLM) with a textual memory mechanism to ensure continuity and contextual coherence. In addition, we introduce a novel evaluation metric, VAD-Score (Video Automated Description Score), to assess precision, recall, and F1 scores for events, subjects, and objects. Our approach delivers competitive results on a diverse set of videos from the DREAM-1K dataset, spanning categories such as live-action, animation, shorts, stock, and YouTube, with a focus on fine-grained comprehension.

  • Open Access

    Perceptual cue-guided adaptive image downscaling for enhanced semantic segmentation on large document images

    by Chulwoo Pack, Leen-Kiat Soh, Elizabeth Lorang

    Image downscaling is an essential operation to reduce spatial complexity for various applications and is becoming increasingly important due to the growing number of solutions that rely on memory-intensive approaches, such as applying deep convolutional neural networks to semantic segmentation tasks on large images. Although conventional content-independent image downscaling can efficiently reduce complexity, it is vulnerable to losing perceptual details, which are important to preserve. Alternatively, existing content-aware downscaling severely distorts spatial structure and is not effectively applicable for segmentation tasks involving document images. In this paper, we propose a novel image downscaling approach that combines the strengths of both content-independent and content-aware strategies. The approach limits the sampling space per the content-independent strategy, adaptively relocating such sampled pixel points, and amplifying their intensities based on the local gradient and texture via the content-aware strategy. To demonstrate its effectiveness, we plug our adaptive downscaling method into a deep learning-based document image segmentation pipeline and evaluate the performance improvement. We perform the evaluation on three publicly available historical newspaper digital collections with differences in quality and quantity, comparing our method with one widely used downscaling method, Lanczos. We further demonstrate the robustness of the proposed method by using three different training scenarios: stand-alone, image-pyramid, and augmentation. The results show that training a deep convolutional neural network using images generated by the proposed method outperforms Lanczos, which relies on only content-independent strategies.


  • Open Access

    Augmentation-based Pseudo-Ground truth Generation for Deep Learning in Historical Document Segmentation for Greater Levels of Archival Description and Access

    by Chulwoo Pack, Yi Liu, Leen-Kiat Soh, Elizabeth Lorang

    The successful use of deep learning solutions for document image segmentation typically relies on a large number of manually labeled ground truth examples, which is expensive to obtain for historical document images that have significant noise effects and variation. At the same time, successful applications of deep learning solutions for document image segmentation have rich potential to facilitate greater levels of description in archival collections (e.g., at and below the item-level). These greater levels of description are critical to increasing access and use of archival collections across an array of research domains. In response, this article investigates whether an augmentation-based approach to generating pseudo-ground truth can be effective with a limited number of labeled images in a document segmentation application. The rationale is that if we can decrease the cost of generating ground truth through augmentation-based approaches, we can use these approaches as part of the description and access pipelines for historical library and archival collections. In this initial exploration, we first generate synthetic images and corresponding pseudo-ground truth using a set of existing degradation-based augmentation models from a small number of labeled actual images. When generating synthetic images, we control the visual quality distortion based on OCR word-level confidence to avoid generating images unlikely to be present in the dataset. Then, we perform several investigations to examine the impact of incorporating pseudo-ground truth data in the training of the deep learning network dhSegment and further evaluate the use of multiple combinations of degradation models. We also assess the generalizability of the approach by applying the trained network on a larger dataset. Our investigations primarily use real-world datasets known to have significant noise effects. Results show that augmentation-based pseudo-ground truth generation is capable of improving segmentation performance with the use of the full original dataset and requires only 30% of the original dataset. Results also show that using more than three degradation models is likely to cause overfitting during training. Furthermore, we show that a segmentation network trained on pseudo-ground truth data has generalization capability.