报告题目：Weakly Supervised Segmentation on 3D Point Clouds
报告人：新加坡南洋理工大学Asst Prof Guosheng Lin
报告时间：2022年11月17日（周四）下午  14:30点
报告地点：腾讯会议（会议号：973-372-282）
主持人：李天瑞教授、吕凤毛副教授

报告摘要：

       In this talk, I will present our recent work on weakly-supervised segmentation methods for 3D point clouds.  Weakly supervised point cloud segmentation, i.e. semantically segmenting a point cloud with only a few labelled points in the whole 3D scene, is highly desirable due to the heavy burden of collecting abundant dense annotations for the model training. However, existing methods remain challenging to accurately segment 3D point clouds since limited annotated data may lead to insufficient guidance for label propagation to unlabeled data. Considering the consistency-based methods have achieved promising progress,  we propose a novel model for weakly supervised point cloud segmentation, where the dual adaptive transformations are performed via an adversarial strategy at both point-level and region-level, aiming at enforcing the local and structural smoothness constraints on 3D point clouds.  Extensive experiments demonstrate that our model can effectively leverage unlabeled 3D points and achieve significant performance gains on public datasets.

报告人简介：
       Guosheng Lin is an Assistant Professor at the School of Computer Science and Engineering, Nanyang Technological University, since 2017. Prior to that, he was a research fellow at the Australian Centre for Robotic Vision from 2014 to 2017. He received his PhD from the University of Adelaide in 2014. His research interests generally lie in deep learning and 2D/3D visual understanding. He has published over 100 research articles in top-tier research venues and his research work receives over 10,000 citations as per Google scholar statistics. He has been serving as a reviewer or meta-reviewer in a number of premier research conferences and journals.