Open-world Segmentation and Tracking in 3D

Explore a Stanford University seminar presentation where NVIDIA researcher Laura Leal-Taixé delves into innovative approaches for training detection and segmentation models using 3D data like Lidar, without requiring direct 3D domain annotations. Learn about various pseudo-labeling techniques and witness a demonstration of Segment Anything in Lidar technology. This 54-minute talk, part of the AA289 Stanford Robotics and Autonomous Systems Seminar series, offers valuable insights into advancing 3D object detection and tracking capabilities for autonomous systems and robotics applications.