Yuyang Zhao

I am currently a Ph.D. candidate in the Computer Vision and Robotic Perception (CVRP) Laboratory at National University of Singapore, under the supervision of A/P Gim Hee Lee. I received my B.E. degree from Tianjin University in 2020.

My research interests lie in building generalizable computer vision systems, especially securing model performance under distribution and category shifts. Currently, I am focusing on generalizable semantic segmentation.


           

profile photo

News
  • [January 2023] I received the Research Achievement Award by NUS!
  • [September 2022] Our AdvStyle is accepted to NeurIPS 2022!
  • [July 2022] I received the Outstanding Reviewer Award in ICML 2022 (Top 10%) !
  • [July 2022] One paper about domain generalized semantic segmentation is accepted to ECCV 2022!
  • [May 2022] One paper about open compound domain adaptation is accepted to IEEE TCSVT!
  • [March 2022] One paper about novel class discovery is accepted to CVPR 2022!
  • [November 2021] One paper about optical flow estimation is accepted to Neurocomputing.
  • [September 2021] One paper about optical flow estimation is accepted to Signal Processing: Image Communication.
  • [March 2021] One paper about domain generalized person re-identification is accepted to CVPR 2021!

Preprints
Style-Hallucinated Dual Consistency Learning: A Unified Framework for Visual Domain Generalization
Yuyang Zhao, Zhun Zhong, Na Zhao, Nicu Sebe, Gim Hee Lee
Arxiv

Extension of our ECCV 2022 paper (SHADE). This paper applies SHADE to visual domain generalization tasks, including semantic segmentation with Transformer backbone, image classification, and object detection.

Synthetic-to-Real Domain Generalized Semantic Segmentation for 3D Indoor Point Clouds
Yuyang Zhao, Na Zhao, Gim Hee Lee
Arxiv
PDF / Code (Coming Soon!)

The first work on domain generalized semantic segmentation in 3D indoor scenes.

Publications

2022

Adversarial Style Augmentation for Domain Generalized Urban-Scene Segmentation
Zhun Zhong*, Yuyang Zhao*, Gim Hee Lee, Nicu Sebe
(* Equal contribution)
NeurIPS, 2022
PDF / Code

AdvStyle adversarially changes the channel-wise mean and standard deviation to diversify source samples.

Style-Hallucinated Dual Consistency Learning for Domain Generalized Semantic Segmentation
Yuyang Zhao, Zhun Zhong, Na Zhao, Nicu Sebe, Gim Hee Lee
ECCV, 2022

We introduce a dual consistency learning framework for domain generalized semantic segmentation, and propose a style hallucination module to generate pair-wise stylized samples.

Novel Class Discovery in Semantic Segmentation
Yuyang Zhao, Zhun Zhong, Nicu Sebe, Gim Hee Lee
CVPR, 2022

The first work focuses on novel class discovery in semantic segmentation. This work addresses the co-occurrence of base, novel and background classes.

Source-Free Open Compound Domain Adaptation in Semantic Segmentation
Yuyang Zhao*, Zhun Zhong*, Zhiming Luo, Gim Hee Lee, Nicu Sebe
(* Equal contribution)
IEEE Transactions on Circuits and Systems for Video Technology, 2022
PDF / Code

2021

Learning to Generalize Unseen Domains via Memory-based Multi-Source Meta-Learning for Person Re-Identification
Yuyang Zhao*, Zhun Zhong*, Fengxiang Yang, Zhiming Luo, Shaozi Li, Nicu Sebe
(* Equal contribution)
CVPR, 2021
PDF / Code
FPCR-Net: Feature Pyramidal Correlation and Residual Reconstruction for Optical Flow Estimation
Xiaolin Song, Yuyang Zhao, Jingyu Yang, Cuiling Lan, Wenjun Zeng
Neurocomputing, 2021
PDF
Spatio-temporal Context-aware Optical Flow Estimation
Xiaolin Song, Yuyang Zhao, Jingyu Yang
Signal Processing: Image Communication, 2021
PDF

Professional Service
  • Program Comittee / Conference Reviewer: CVPR 2022-2023, ECCV 2022, ICML 2022, NeurIPS 2022, AAAI 2022, ACM MM 2022
  • Journal Reviewer: IEEE Transactions on Multimedia

Awards
  • Research Achievement Award, National University of Singapore, 2023
  • Outstanding Reviewer Award, ICML, 2022
  • Research Scholarship, National University of Singapore, 2021


Stolen from Jon Barron