Wonseok Oh 오원석

I'm a first-year Ph.D. student at the University of Colorado Boulder, advised by Prof. Alan Bovik, working on generative video, visual quality, and computational perception. My Ph.D. is funded by Meta (Reality Labs).

I also work with Alex Chapiro at Meta, and closely collaborate with Prof. João Henriques at the Oxford VGG lab. Previously, I completed my M.S. at the University of Michigan, Ann Arbor, where I was fortunate to be advised by Prof. Andrew Owens. I received a B.S. from Korea University in 2022 (double major in Computer Science & Engineering and Chemical & Biological Engineering).

profile photo
News
2026
Our paper "Sudo Thermal Camera" is accepted to the CVPR 2026 Workshop.
2026
Nominated as a Gold Reviewer for ICML 2026.
2025
Received Ph.D. funding from Meta Reality Labs.
2025
Began my Ph.D. under the supervision of Prof. Alan Bovik at CU Boulder.
2025
Our paper "Hearing Hands" is accepted to CVPR 2025.
Research Interests

My goal is to build generative models of the visual world that are physically grounded, perceptually faithful, and useful to act on. I approach this from four angles that feed into one another:

  • Generative Video: physically-grounded, controllable video synthesis from a single image guided by simulation, so that generation obeys the dynamics of the real world such as heat, motion, and material behavior.
  • Perceptual Quality: subjective and objective quality assessment of rendered and generated content that turns what people actually perceive into a signal that systems can be measured against and optimize for.
  • Multimodal Perception: connecting vision with other modalities such as sound, touch, and language to model the physical world beyond what any single sense reveals.
  • Vision for Robotics: language and image-conditioned learning that turns perceptual understanding into action so that embodied agents can interpret scenes and act in real-world environments.

    • Publications ( / )

    (* indicates equal contribution)

    Sudo Thermal Camera: Heat-Adjustable Generative Video from a Single RGB Image
    Wonseok Oh, Alan Bovik
    CVPR Workshop 2026 · Under review
    paper

    A training-free system that takes a single RGB image and a user-specified heat-source map to generate a physically plausible video of the scene under that heat.
    Time is the Reward: Learning Vision-Language Reward Potentials for Robotics
    Under review

    A simple time-indexed potential-shaping signal that turns vision-language progress into dense rewards for robot manipulation.
    Hearing Hands: Generating Sounds from Physical Interactions in 3D Scenes
    Yiming Dou, Wonseok Oh, Yuqing Luo, Antonio Loquercio, Andrew Owens
    CVPR 2025
    paper · project page · code

    We make 3D scene reconstruction interactive by predicting the sounds of human hands physically interacting with the scene.
    From 2D Portraits to 3D Realities: Advancing GAN Inversion for Enhanced Image Synthesis
    Wonseok Oh*, Youngjoo Jo*
    CVPR 2024 Workshop on Learning 3D with Multi-View Supervision (Oral)
    paper
    Improving the Quality of Pixel-wise Transfer using Attention-based Encoding (SAE) and an Abortion Method
    Wonseok Oh et al.
    Korea Computer Congress 2023 (Oral), 2nd award in independent study, Korea University
    paper
    Visualization Comparison of GANs for Reconstructing a De-identified Image Dataset using Attention
    Wonseok Oh, Kangmin Bae, Yuseok Bae
    Korea Software Congress 2021
    html
    RDID-GAN: Reconstructing a De-identified Image Dataset to Generate Effective Learning Data
    Wonseok Oh, Kangmin Bae, Yuseok Bae
    Journal of KIISE, 2021
    html
    Service
    • Conference reviewer: ICML 2026, ICLR 2025, CPAL 2025
    • Workshop reviewer: CVPRW 2024 (Generative Models for Computer Vision; AI City Challenge), ECCVW 2024 (Out-of-Distribution Generalization in Computer Vision; AI for Visual Arts)
    Education
    University of Colorado Boulder
    2025.08 — Present
    Boulder, U.S.
    Ph.D. in Electrical & Computer Engineering
    Advisor: Prof. Alan Bovik
    University of Michigan
    2023.08 — 2025.05
    Ann Arbor, U.S.
    M.S. in Electrical & Computer Engineering
    Advisor: Prof. Andrew Owens
    Korea University
    2015.03 — 2023.08
    Seoul, South Korea
    B.S. in Computer Science and Engineering & B.E. in Chemical & Biological Engineering
    Experience
    Meta Reality Labs
    2025.08 — Present
    Research Assistant
    Supervisor: Alex Chapiro
    VGG, University of Oxford
    2025.06 — Present
    Research Collaboration
    Supervisor: Prof. João Henriques
    ETRI
    2021.07–08, 2022.01–02, 2022.07–08
    Daejeon, South Korea
    Research Assistant
    With Kimin Yun and Yongju Lee
    Selected Honors
    • Gold Reviewer, ICML 2026
    • 2nd Award, Independent Study (KCC 2023 oral), Korea University, 2023

    Template from Jon Barron's website