Seokyeon Kim

Seokyeon Kim

Auto-generated profile, based on publicly available information

Seokyeon Kim

Computer Engineering and Convergence Engineering for Intelligent Drone, Sejong University, Seoul 05006, Korea

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 July 2021 in Sensors
Verified
Journal Article
Open access
PDF file
SYSangbong Yoo
Sangbong Yoo
SJSeongmin Jeong
Seongmin Jeong
SKSeokyeon Kim
Seokyeon Kim
Yun JangYun Jang
Yun Jang
Gaze movement and visual stimuli have been utilized to analyze human visual attention intuitively. Gaze behavior studies mainly show statistical analyses of eye movements and human visual attention. During these analyses, eye movement data and the saliency map are presented to the analysts as separate views or merged views. However, the analysts become frustrated when they need to memorize all of the separate views or when the eye movements obscure the saliency map in the merged views. Therefore, it is not easy to analyze how visual stimuli affect gaze movements since existing techniques focus excessively on the eye movement data. In this paper, we propose a novel visualization technique for analyzing gaze behavior using saliency features as visual clues to express the visual attention of an observer. The visual clues that represent visual attention are analyzed to reveal which saliency features are prominent for the visual stimulus analysis. We visualize the gaze data with the saliency features to interpret the visual attention. We analyze the gaze behavior with the proposed visualization to evaluate that our approach to embedding saliency features within the visualization supports us to understand the visual attention of an observer.

24 June 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
HSHyeSook Son
HyeSook Son
SKSeokyeon Kim
Seokyeon Kim
HYHanbyul Yeon
Hanbyul Yeon
YKYejin Kim
Yejin Kim
Yun JangYun Jang
Yun Jang
SKSeung-Eock Kim
Seung-Eock Kim
The output of a deep-learning model delivers different predictions depending on the input of the deep learning model. In particular, the input characteristics might affect the output of a deep learning model. When predicting data that are measured with sensors in multiple locations, it is necessary to train a deep learning model with spatiotemporal characteristics of the data. Additionally, since not all of the data measured together result in increasing the accuracy of the deep learning model, we need to utilize the correlation characteristics between the data features. However, it is difficult to interpret the deep learning output, depending on the input characteristics. Therefore, it is necessary to analyze how the input characteristics affect prediction results to interpret deep learning models. In this paper, we propose a visualization system to analyze deep learning models with air pollution data. The proposed system visualizes the predictions according to the input characteristics. The input characteristics include space-time and data features, and we apply temporal prediction networks, including gated recurrent units (GRU), long short term memory (LSTM), and spatiotemporal prediction networks (convolutional LSTM) as deep learning models. We interpret the output according to the characteristics of input to show the effectiveness of the system.

21 October 2024 in IEEE Transactions on Visualization and Computer Graphics
Verified
Journal Article
Closed access
SYSangbong Yoo
Sangbong Yoo
SKSeokyeon Kim
Seokyeon Kim
Yun JangYun Jang
Yun Jang
The transfer function (TF) design is crucial for enhancing the visualization quality and understanding of volume data in volume rendering. Recent research has proposed various multidimensional TFs to utilize diverse attributes extracted from volume data for controlling individual voxel rendering. Although multidimensional TFs enhance the ability to segregate data, manipulating various attributes for the rendering is cumbersome. In contrast, low-dimensional TFs are more beneficial as they are easier to manage, but separating volume data during rendering is problematic. This paper proposes a novel approach, a two-level transfer function, for rendering volume data by reducing TF dimensions. The proposed technique involves extracting multidimensional TF attributes from volume data and applying t-Stochastic Neighbor Embedding (t-SNE) to the TF attributes for dimensionality reduction. The two-level transfer function combines the classical 2D TF and t-SNE TF in the conventional direct volume rendering pipeline. The proposed approach is evaluated by comparing segments in t-SNE TF and rendering images using various volume datasets. The results of this study demonstrate that the proposed approach can effectively allow us to manipulate multidimensional attributes easily while maintaining high visualization quality in volume rendering.

20 June 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Yun JangYun Jang
Yun Jang
SKSeokyeon Kim
Seokyeon Kim
Understanding and perceiving three-dimensional scientific visualizations, such as volume rendering, benefit from visual cues produced by the shading models. The conventional approaches are local shading models since they are computationally inexpensive and straightforward to implement. However, the local shading models do not always provide proper visual cues since non-local information is not sufficiently taken into account for the shading. Global illumination models achieve better visual cues, but they are often computationally expensive. It has been shown that alternative illumination models, such as ambient occlusion, multidirectional shading, and shadows, provide decent perceptual cues. Although these models improve upon local shading models, they still require expensive preprocessing, extra GPU memory, and a high computational cost, which cause a lack of interactivity during the transfer function manipulations and light position changes. In this paper, we proposed an approximate image-space multidirectional occlusion shading model for the volume rendering. Our model was computationally less expensive compared to the global illumination models and did not require preprocessing. Moreover, interactive transfer function manipulations and light position changes were achievable. Our model simulated a wide range of shading behaviors, such as ambient occlusion and soft and hard shadows, and can be effortlessly applied to existing rendering systems such as direct volume rendering. We showed that the suggested model enhanced the visual cues with modest computational costs.

30 November 2021 in Journal of the Korea Computer Graphics Society
Verified
Journal Article
Open access
PDF file
SKSeokyeon Kim
Seokyeon Kim
Yun JangYun Jang
Yun Jang
사단법인 한국컴퓨터그래픽스학회의 논문지로서 그래픽스 분야와 그에 연관된 연구영역의 새로운 발견과 최첨단 연구결과를 발표하는 토론의 장입니다. 학회를 대표하는 학술지인 논문지는 1995년에 논문지 발간을 시작으로 2000년부터 3월, 6월, 9월, 12월 총 4회 발간되었고, 2015년부터는 KCGS 학술대회 특별호를 7월에 발간하면서 총 5회 발간되고 있습니다.

30 April 2021 in Journal of KIISE
Verified
Journal Article
Closed access
YKYejin Kim
Yejin Kim
SKSeokyeon Kim
Seokyeon Kim
Yun JangYun Jang
Yun Jang
COVID-19는 감염자의 비말을 통해 전파된다. 비말의 전파는 시공간의 영향을 받는다. 전염병의 전파는 감염자와 비감염자의 건강 상태, 환경적 요인 등 다양한 요인의 상호작용으로 이루어진다. 하지만 예측 모델에 전염병과 관련된 정보를 모두 포함하고, 정보간의 관계를 파악하는 것은 쉽지 않다. 본 논문에서는 COVID-19의 전염 특징을 딥러닝 학습 데이터셋에 포함하고, COVID-19 확산 데이터의 조합이 딥러닝 예측 성능에 미치는 영향을 파악하는 연구 방법을 제안하였다. 예측에 앞서 COVID-19의 전염 특징을 파악하고, 데이터 전처리 시 COVID-19 확산 특징을 포함하기 위한 고려 사항을 정의하였다. 딥러닝 모델링 시에는 시공간 예측을 위해 ConvLSTM을 응용한 예측 모델을 설계하였다. 예측 모델을 테스트하는 단계에서는 확산 데이터를 여러 가지 방식으로 조합하고, 각 조합이 딥러닝 예측 성능에 미치는 영향을 분석하였다. 성능 평가를 위해 COVID-19 확진자의 정보와 확진자가 방문한 장소의 특징을 기준으로 47개의 피처를 만들고, 120개의 피처 조합을 실험하였다. 또한 모델 성능 평가 지표로 MAPE를 이용하였다. 실험 결과, COVID-19 데이터셋에서 피처 조합 모델의 MAPE 평균값으로 1.234, 피처를 조합하지 않은 모델의 MAPE 평균값으로 2.217을 얻을 수 있었다.