Taesung Kwon

Taesung Kwon

Auto-generated profile, based on publicly available information

Taesung Kwon

Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

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

Representative Publications

View All Publications
19 November 2021 in IEEE Transactions on Computational Imaging
Verified
Journal Article
Closed access
TKTaesung Kwon
Taesung Kwon
JYJong Chul Ye
Jong Chul Ye
Recently, CycleGAN was shown to provide high-performance, ultra-fast denoising for low-dose X-ray computed tomography (CT) without the need for a paired training dataset. Although this was possible thanks to cycle consistency, CycleGAN requires two generators and two discriminators to enforce cycle consistency, demanding significant GPU resources and technical skills for training. A recent proposal of switchable CycleGAN with Adaptive Instance Normalization (AdaIN) alleviates the problem in part by using a single generator. However, two discriminators and an additional AdaIN code generator are still required for training. To solve this problem, here we present a novel cycle-free CycleGAN architecture, which consists of a single generator and a discriminator but still guarantees the cycle consistency. The main innovation comes from the observation that the use of an invertible generator automatically fulfills the cycle consistency condition and eliminates the additional discriminator in the CycleGAN formulation. Extensive experiments using various levels of low-dose CT images confirm that our method can improve denoising performance using only 24% of learnable parameters, twice faster training time, and showed better trade-offs between the complexity versus reconstruction quality, compared to the existing methods.

1 September 2013 in Journal of Asia-Pacific Entomology
Verified
Journal Article
Closed access
TKTae-Sung Kwon
Tae-Sung Kwon
YPYoung Kyu Park
Young Kyu Park
JLJoo-Hoon Lim
Joo-Hoon Lim
SRSae Han Ryou
Sae Han Ryou
CLCheol Min Lee
Cheol Min Lee

1 April 2021 in Global Ecology and Conservation
Verified
Journal Article
Open access
TKTae-Sung Kwon
Tae-Sung Kwon
CLCheol Min Lee
Cheol Min Lee
EKEun-Sook Kim
Eun-Sook Kim
MWMyoungsoo Won
Myoungsoo Won
SKSung-Soo Kim
Sung-Soo Kim
Young-Seuk ParkYoung-Seuk Park
Young-Seuk Park

13 August 2025 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
TKTaesung Kwon
Taesung Kwon
GSGookho Song
Gookho Song
YKYoosun Kim
Yoosun Kim
JKJeongsol Kim
Jeongsol Kim
JYJong Chul Ye
Jong Chul Ye
MJMooseok Jang
Mooseok Jang
Imaging through scattering is challenging, as even a thin layer can randomly perturb light propagation and obscure hidden objects. Accurate closed-form modeling of forward scattering remains difficult, particularly for dynamically varying or thick layers. Here, we introduce a plug-and-play inverse solver based on video diffusion models with a physically grounded forward model tailored to dynamic scattering layers. Our method extends Diffusion Posterior Sampling (DPS) to the spatio-temporal domain, thereby capturing statistical correlations between video frames and scattered signals more effectively. Leveraging these temporal correlations, our approach recovers high-resolution spatial details that spatial-only methods typically fail to reconstruct. We also propose an inference-time optimization with a lightweight mapping network, enabling joint estimation of low-dimensional forward-model parameters without additional training. This joint optimization significantly enhances adaptability to unknown, time-varying degradations, making our method suitable for blind inverse scattering problems. We validate across diverse conditions, including different scene types, layer thicknesses, and scene-layer distances. And real-world experiments using multiple datasets confirm the robustness and effectiveness of our approach, even under real noise and forward-model approximation mismatches. Finally, we validate our method as a general video-restoration framework across dehazing, deblurring, inpainting, and blind restoration under complex optical aberrations.

29 July 2021 in Antimicrobial Stewardship & Healthcare Epidemiology
Verified
Journal Article
Open access
PDF file
TKTaesung Kwon
Taesung Kwon
SKStacy Kenyon
Stacy Kenyon
KKKimberly Kilheeney
Kimberly Kilheeney
SMStanley Martin
Stanley Martin
MSMark Shelly
Mark Shelly
Background: Determining the incidence of SARS-CoV-2 in healthcare workers (HCW) is important in assessing the safety of the work environment. Though of limited use in acute illness, serologic testing can detect some infections that occur undetected. We compared the prevalence of antibodies to SARS-CoV-2 to a place of work, exposure by role and department, and use of various prevention methods. Methods: Healthcare workers (HCWs) working in Geisinger Health System were offered voluntary serology through Employee Health. Before they had blood taken, they completed a brief questionnaire. Testing was conducted from June 15 to September 4, 2020. Blood was analyzed for SARS-CoV-2 immunoglobulin G (IgG) (Roche and Diasorin platforms). Results: In total, 2,295 employees and contract workers providing care at Geisinger facilities were tested. Most of this group, 2,037 (88.8%), were involved in direct patient care. In total, 101 tests returned positive, a rate of 4.4% (95% CI, 3.6%–5.3%). Of 54 HCWs with a positive NAAT for SARS-CoV-2, positive serology results were found in 48, a sensitivity of 89% (95% CI, 78%–95%). Those involved in patient care were slightly more likely to become infected, 91 of 2,037 (4.6%) compared to 10 of 258 who were not involved in patient care (3.9%; P = .68). Those with unprotected exposure to a known case of COVID-19 were more likely than those not exposed to be positive for SARS-CoV-2, 51 of 792 (6.4% vs 3.3%; P = .0008). This risk was highest for those exposed outside work (7 of 33; 21%; P = .003). HCWs working in COVID-19 units were positive at a rate of 4.0% (95% CI, 3.8%–5.4%), no more than other inpatient areas, which were 5.0% positive (95% CI, 3.8%–6.4%). HCWs working with outpatients were at slightly lower risk, 2.8% positivity (95% CI, 1.9%–4.1%). The rates of infection ranged between 3.3% and 5.0% by job category. Employees were asked about symptoms experienced since March 2020. Positive serology occurred in 39 (2.8%) of 1,414 employees who did not recall any symptoms. Symptoms related to COVID-19, except sore throat, were strongly correlated with positive serology. Conclusions: When provided a safe work environment, the risk of COVID-19 in employees is comparable to that in the surrounding communities. Persons with patient care responsibilities have an absolute risk that is marginally higher.Funding: NoDisclosures: None