Yu Kong

Dr. Yu Kong is an Assistant Professor of the Department of Computer Science and Engineering, Michigan State University. He is now directing the ACTION Lab. He serves Pattern Recognition and IEEE Transaction as an Associate Editor. He also serves CVPR and ICCV as an Area Chair. His research has been supported by NSF and DoD. He is an IEEE Senior Member.

Dr. Kong is interested in video understanding, vision-language modeling, video diffusion, open world vision, and their applications in Embodied AI, Biometrics, and Cybersecurity. More specifically, he works on action recognition, visual forecasting, video question-answering, video grounding, open-set recognition, and meta-learning.

Openings: We have openings for Fall 2025 for Ph.D. students, Visiting Scholars, and Research Assistants. Due to large volume of emails, I am not able to track and respond to every applicant. Please fill in the external contact form, and I will review applications periodically. How to apply?

Research Areas

• Video Understanding
  • Action Recognition: A survey [IJCV-22], DEAR: Deep Evidential Action Recognition [ICCV-21], Max-Margin Heterogenous Information Machine [IJCV 2017]
  • Action Detection: Open vocabulary action detection [WACV 2025], GateHUB [CVPR 2022], OpenTAL: open-set temporal action localization [CVPR 2022]
  • Visual Forecasting: Egocentric 3D Trajectory Forecasting [ICCV-23], Adversarial Action Prediction Networks [T-PAMI 2020], DRIVE [ICCV 2021], Group Activity Prediction [ECCV 2020], Uncertainty-based Traffic Accident Anticipation [MM 2020]
• Vision-language Modeling: Visual Reasoning [MM-23], Video Entailment [ICCV-21], Visual Grounding [MM-20]
• Open-world Recognition: Open-set Action Recognition [ICCV 2021], Open-set Action Localization [CVPR 2022], Generalized Open-set Recognition [WACV-23], Compositional Grounding [ECCV-24], Compositional Zero-shot Learning [ECCV-24]

[Recent work:] survey on visual large language model, instructional video generation, action analysis, and vision foundation models.