Description
3D point clouds are unordered and sparse, and early methods converted them to voxels or images, losing precision. PointNet from Stanford is the pioneering deep learning model that takes raw point clouds directly for 3D classification and segmentation.
It laid the foundation for point cloud deep learning, widely used in autonomous driving, robotics and 3D vision.
Raw points:No conversion.
Tasks:Classification and segmentation.
Foundational:Field-defining.
It laid the foundation for point cloud deep learning, widely used in autonomous driving, robotics and 3D vision.
Features
Raw points:No conversion.
Tasks:Classification and segmentation.
Foundational:Field-defining.
