About me

I am a PhD student at Seoul National University, advised by Prof. Gunhee Kim. Before joining SNU, I received MSc in School of Computing at KAIST, supervised by Prof. Seunghoon Hong. Also, I majored Electrical and Electronic Engineering at Hanyang University (HYU), from which I obtained my B.S. degree.

My research interests lie in the machine learning and computer vision, where I love to tackle challenging problems through generative modelling and representation learning. In particular, I worked on a video prediction project with wonderful researchers at Google Brain where we showed unprecedentedly high-quality and long-term video predictions are possible through our hierarchical generative model. Prior to that, I led and participated in projects related to unsupervised disentangled representation learning, where a well-known challenge is to represent and synthesize data by discovering independent factors of variations in data through generative models.

During my MS program, I worked as an AI research scientist at ART Lab, where I led a project for digitizing and interpreting human faces by solving multi-task learning problems, including 3D reconstruction, semantic segmentation, object detection, landmark detection, pose estimation, attributes regression, some of which were presented in Consumer Technology Association (CES) 2021. While working at ART Lab, I also contributed to one patent application which is related to methodology and system for automatic photography using electronic devices.

Publications

[ICLR’22] Multi-Task Processes.
Donggyun Kim, Seongwoong Cho, Wonkwang Lee, Seunghoon Hong.
[Paper] [Code]

[ICLR’21] Revisiting Hierarchical Approach for Persistent Long-Term Video Prediction.
Wonkwang Lee, Whie Jung, Han Zhang, Ting Chen, Jing Yu Koh, Thomas Huang, Hyungsuk Yoon, Honglak Lee, Seunghoon Hong.
[Paper] [Code] [Project Page]

[AAAI’21] IB-GAN: Disentangled Representation Learning with Information Bottleneck GAN.
Insu Jeon, Wonkwang Lee, Myeongjang Pyeon, Gunhee Kim,
[Paper]

[ECCV’20] High-Fidelity Synthesis with Disentangled Representation.
Wonkwang Lee, Donggyun Kim, Seunghoon Hong, Honglak Lee.
[Paper] [Code]