Title
Adverse Weather Visual Restoration: From Classical CNNs to Foundation Models
Abstract
In this talk, I will present our works for adverse weather image/video restoration by leveraging diverse deep learning models. First, I will focus on leveraging traditional CNNs (e.g., Transformer) for addressing different adverse weather image/video restoration tasks. Then, I will talk about our recent works based on diffusion models and LLMs.
Biographies
Dr. Lei ZHU is currently working as an Assistant Professor at Robotics and Autonomous Systems (ROAS) Thrust & DSA Thrust at HKUST (Guangzhou). He received his PhD degree at CSE of the CUHK in 2017, and then worked as a postdoctoral researcher at University of Cambridge. His research interest is to develop intelligent visual perception theory and algorithms to assist autonomous systems, medical imaging, robotics, and so on. In the past few years, he has published 100+ papers in top-tier conferences or journals (e.g., IEEE TPAMI/IJCV/NeurIPS/CVPR/ICCV). He has been invited as a program chair for ACM SIGGRAPH VRCAI 2022 & 2024, and an area chair for ICLR 2025, CVPR 2025, ECCV 2024, MICCAI 2024 & 2023, and so on.
