Seong Joon Oh

Seong Joon Oh is a faculty member in the Department of Computer Science at KAIST. His research is associated with the Kim Jaechul Graduate School of AI. He can be contacted via email at coallaoh@gmail.com, and his professional profile is available at https://engineering.kaist.ac.kr/people/seongjoonoh.com.

Seong Joon Oh is a researcher specializing in machine learning and computer vision. His work focuses on improving the robustness and reliability of machine learning models, particularly in areas such as regularization strategies, generative models, and privacy protection. He has contributed to the development of techniques like CutMix for training strong classifiers and has explored the challenges in scene text recognition and model comparisons. His research also addresses issues like adversarial perturbations, privacy leakage, and the evaluation of weakly supervised methods. Oh's work emphasizes the importance of fidelity and diversity metrics in generative models and the need for de-biased representations in machine learning.