报告人：Yuexin Ma
主持人：Ruizhen Hu
日期：2024.08.09
时间：9:30am
地点：深圳大学致真楼801
Abstract
Recent advancements in foundation models have spurred AI development to new heights. However, constructing 3D vision foundation models is challenging due to the high cost of acquiring and annotating 3D data.
To tackle this, we've approached the problem from two angles. First, we've developed label-efficient learning algorithms tailored for 3D scenes. These algorithms excel in unsupervised learning, domain adaptation, label-free learning, and open vocabulary learning tasks. Second, we've focused on fine-grained understanding for human-centric scenes. We proposed several large-scale datasets and benchmarks for understanding dense crowds, human-object interactions, and human motions. These efforts are significant for building 3D vision foundation models and are crucial for applications like autonomous driving, service robots, and human-robot collaboration.
Bio
Yuexin Ma is an Assistant Professor in SIST, ShanghaiTech University. She received the PhD degree from the University of Hong Kong. Her research interests include computer vision and artificial intelligence. Particularly, her current research focuses on 3D scene understanding, multi-modal learning, autonomous driving, embodied AI, etc.
She has published more than 60 papers on top journals and conferences, including Science Robotics, TPAMI, CVPR, ICCV, ECCV, AAAI, SIGGRAPH, etc., which have obtained more than 3000 citations. Her first-author paper had been awarded as one of the most influential AAAI-19 papers. More information can be found in her homepage (http://yuexinma.me/ ).
