Research on Cross-Domain Semantic Segmentation Based on Unsupervised Domain Adaptation

2024年5月23日 13:00 - 14: 00

信息管理与工程学院308室

Yiting Cheng(程逸婷) obtained her Ph.D degree from the School of Computer Science, Fudan University in 2024, advised by Prof. Wenqiang Zhang. Her research primarily focuses on the semantic segmentation algorithms under diverse environmental conditions. Prior to her ph.D, Yiting completed her graduate studies from Fudan University in 2021. She Received the Outstanding Student Award from Fudan University in 2024 and the Shanghai Outstanding Student Award in 2021.

Research on Cross-Domain Semantic Segmentation Based on Unsupervised Domain Adaptation

Robust environmental perception stands as the cornerstone for the interaction of intelligent agents with the real world, facilitating the accurate decision-making necessary for various applications such as autonomous driving, robot navigation, and virtual reality. To realize this goal, the development and exploration of semantic segmentation algorithms capable of flexible generalization under diverse environmental conditions become imperative. Recently, propelled by advancements in algorithms, computing power, and data availability, deep learning models have showcased efficient and accurate semantic segmentation capabilities within fixed environment. However, real-world applications often entail complex and dynamic environmental changes, posing a substantial challenge to cross-domain semantic segmentation when confronted with shifts in data distributions. This report will primarily introduce my research and progress on enhancing cross-domain semantic segmentation generalization using unsupervised domain adaptation algorithms.