Donggoo Jung

I am a PhD student at the Visual Intelligence Lab., part of Department of Artificial Intelligence at Hanyang University in Seoul, South Korea, where I work on computer vision. My Ph.D. advisor is Tae Hyun Kim.

Before starting my Ph.D. course, I completed Naver's Software Membership program from December 2013 to June 2014 and interned at Naver's data center. From 2014 to 2019, I worked at LG Electronics, primarily developing VR, AR, and computer vision AI applications. Additionally, from 2019 to 2021, I worked at Hyundai Motor Company as an engineer responsible for computer vision-based automated inspection solutions.

I am primarily interested in addressing noise and color distortion that occur in under-exposed and over-exposed conditions, particularly during post-capture processing. (Image Denosing, Low-Light Image Enhancement, Exposure Correction, etc.) I am also working on research in the field of AI-based medical diagnosis.

GitHub  /  Google Scholar  /  LinkedIn  /  Research Logs

News

• [Feb, 2025] One paper is accepted to CVPR 2025!
• [Jan, 2025] One paper is accepted to ICLR 2025, selected as Spotlight paper!
• [Sep, 2024] Invited to Toronto Metropolitan University as a visiting Ph.D. student! (Advisor: Guanghui Wang)
• [Jan, 2024] One paper is accepted to ICLR 2024!

Academic Service

• Reviewer of CVPR (2024,2025), ICCV (2025) and ICLR (2025)

Research (Conference)

* indicates equal contribution.

Exposure-slot: Exposure-centric representations learning with Slot-in-Slot Attention for Region-aware Exposure Correction

We introduce a slot-in-slot mechanism for exposure correction, hierarchically partitioning exposure-centric features, and achieved state-of-the-art performance in the exposure correction task.

Continuous Exposure Learning for Low-light Image Enhancement using Neural ODEs

We are the first to reformulate a curve adjustment-based method using Neural ODEs, achieving unsupervised state-of-the-art performance in the low-light image enhancement task.

sRGB Real Noise Modeling via Noise-Aware Sampling with Normalizing Flows

We leverage the mean and standard deviation of Normalizing Flow to effectively model diverse noise distributions with a single model, achieving state-of-the-art performance in noise modeling in terms of both KLD/AKLD and parameter efficiency.

Research (Journal)

* indicates equal contribution.

Machine learning-based analysis for prediction of surgical necrotizing enterocolitis in very low birth weight infants using perinatal factors: a nationwide cohort study

Learning-based longitudinal prediction models for mortality risk in very-low-birth-weight infants: a nationwide cohort study

Two-stage learning-based prediction of bronchopulmonary dysplasia in very low birth weight infants: a nationwide cohort study

Development of artificial neural networks for early prediction of intestinal perforation in preterm infants

Design and source code from Jon Barron's website