报告题目：Neural Rendering based SLAM
报告人：韩治中 韦恩州立大学（美国）
报告摘要：
Neural rendering has emerged as a powerful tool for novel view synthesis, 3D reconstruction, and generative modeling. By minimizing rendering errors between 3D representations—such as radiance fields that capture both scene geometry and appearance—and 2D image observations, we can reconstruct realistic 3D scenes and render them from unseen viewpoints. This paradigm offers a promising new direction for Simultaneous Localization and Mapping (SLAM), where the goals are to track camera poses, infer 3D geometry, and enable view synthesis simultaneously.
In this talk, I will present recent advances in neural rendering–based SLAM systems, focusing on implicit and explicit radiance field representations such as NeRF and 3D Gaussian Splatting. I will also introduce our recent contributions, which leverage stronger geometric priors, enforce multi-view consistency more effectively, and develop improved tracking and mapping strategies. Together, these innovations lead to more accurate geometry reconstruction, robust camera pose estimation, higher-quality novel view synthesis, and more efficient, scalable SLAM systems.
报告人简介：
Zhizhong Han is an Assistant Professor in the Department of Computer Science at Wayne State University, where he founded and directs the Machine Perception Lab. Prior to joining WSU in 2021, he was a Postdoctoral Researcher with Prof. Matthias Zwicker at the University of Maryland, College Park. His research focuses on 3D computer vision and digital geometry processing. He has served as an Area Chair for premier conferences in computer vision and machine learning, including CVPR, ICCV, ICML, and NeurIPS, and has published extensively at these venues. He was an awardee of the NVIDIA Academic Grant and the Richard Barber Interdisciplinary Research Award.
报告时间：2025年8月24日（星期日）上午10:30
报告地点：吉林大学正新楼三楼报告厅
主办单位：吉林大学人工智能学院
