WKWonsuk Kim

Wonsuk Kim

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

Wonsuk Kim has a total of 13 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Depression Screening Positivity, Treatment-Seeking Intention, and Willingness to Take Antidepressants in a Self-Selected Online Sample of Korean Adults: Cross-Sectional Questionnaire Study

    by Ha-Na Kim, Chul-Min Kim, Young Kyu Park, Kiheon Lee, Youn Huh, Wonsuk Kim, Young Sik Kim

    Background: Depression is a prevalent psychiatric condition with functional and socioeconomic consequences. Understanding the barriers to professional help-seeking and treatment acceptance among individuals with depressive symptoms is important. Objective: This study aimed to quantify depression screening positivity and examine treatment-seeking intention, willingness to use antidepressants, and perceived barriers among Korean adults, with attention to prior depression-related health care experiences. Methods: A cross-sectional online survey was conducted in South Korea (March 16-April 5, 2024) via paid social media advertising and messaging app invitations. A total of 1053 adults aged 20 years and older participated in this study. Depression screening was conducted using 2 Primary Care Evaluation of Mental Disorders–derived items, with the endorsement of either item defining a positive screen. History of depression diagnosis or treatment, intentions toward professional treatment, antidepressant use, and perceived barriers were assessed. Results: Overall, 65% (684/1053) of the participants had a positive depression screen. A positive depression screen was not associated with the intention to seek professional treatment for depression (adjusted odds ratio [aOR] 1.04, 95% CI 0.77‐1.41). Participants with a positive depression screen had lower odds of being willing to take antidepressants than those with a negative screen (aOR 0.73, 95% CI 0.54‐0.99). A within-participant analysis showed that the association of depression screening status differed between treatment-seeking intention and willingness to take antidepressants (ratio of odds ratios 0.70, 95% CI 0.53‐0.94; interaction P =.02). The joint response of intending to seek treatment while being unwilling to take antidepressants was more common among participants with a positive than a negative depression screen (75/384, 11% vs 23/369, 6.2%; aOR 1.70, 95% CI 1.04‐2.77). A history of depression diagnosis or treatment was positively associated with treatment-seeking intention and willingness to use antidepressants. Among screen-positive adults, those without such a history more often reported that they would not seek treatment if diagnosed with depression (129/505, 25.5% vs 26/179, 14.5%; P =.002) or would be unwilling to take antidepressants if recommended (154/505, 30.5% vs 29/179, 16.2%; P <.001). Barrier profiles to treatment seeking differed by history of depression diagnosis or treatment (interaction P <.001), whereas the patterns of reasons for unwillingness to take antidepressants did not differ according to history (interaction P =.08). Conclusions: In this online survey sample, a positive depression screen was associated with a lower willingness to use antidepressants but not with treatment-seeking intention. Among screen-positive participants, those without prior depression-related health care experience had lower help-seeking intention and willingness to take antidepressants than those with prior contact, suggesting the need to reduce barriers to care and support earlier treatment initiation.

  • Open Access

    A Multi-Stage Deep Learning Approach Incorporating Text-Image and Image-Image Comparisons for Cheapfake Detection

    by Jangwon Seo, Hyo-Seok Hwang, Jiyoung Lee, Minhyeok Lee, Wonsuk Kim, Junhee Seok

    The advancement of multimedia and artificial intelligence (AI) technologies has dismantled the barriers of information sharing, yet it has also ushered in a double-edged sword: a surge in the spread of fake information. In this context, there is a growing need for research on the detection of 'cheapfakes,' which are low-cost fake media, known for their ease of creation. This paper proposes a multi-stage deep learning process designed to effectively detect the diverse and rapidly evolving nature of cheapfakes. A single-step deep learning model faces limitations in distinguishing various types of cheapfakes, necessitating the application of a complex deep learning model approach to detect subtle Out-of-Context (OOC) phenomena. This study employs models based on Bidirectional Encoder Representations from Transformers (BERT) and stable diffusion technologies to approach cheapfake detection. Through the ACM ICMR 2024 challenge, the performance of this model was evaluated on a real dataset, achieving an accuracy of 71.9% in Task 1, an improvement of 7% over previous methods, and an accuracy of 55.7% in Task 2. These results are expected to make a significant contribution to the development of strategies for creating and countering cheapfakes. Additionally, this research aims to contribute to the detection of OOC media misuse through this challenge.


  • Indoor Semantic Segmentation for Robot Navigating on Mobile

    by Wonsuk Kim, Junhee Seok

    In recent years, there have been many successes of using Deep Convolutional Neural Networks (DCNNs) in the task of pixel-level classification (also called “semantic image segmentation”). The advances in DCNN have led to the development of autonomous vehicles that can drive with no driver controls by using sensors like camera, LiDAR, etc. In this paper, we propose a practical method to implement autonomous indoor navigation based on semantic image segmentation using state-of-the-art performance model on mobile devices, especially Android devices. We apply a system called `Mobile DeepLabv3', which uses atrous convolution when applying semantic image segmentation by using MobileNetV2 as a network backbone. The ADE20K dataset is used to train our models specific to indoor environments. Since this model is for robot navigating, we re-label 150 classes into 20 classes in order to easily classify obstacles and road. We evaluate the trade-offs between accuracy and computational complexity, as well as actual latency and the number of parameters of the trained models.