Learning Hierarchical Orthogonal Prototypes for Generalized Few-Shot 3D Point Cloud Segmentation

1. 1Learn hierarchical orthogonal prototypes using an entropy-based approach
2. 2Adapt to novel classes with few annotations while preserving base-class performance
3. 3Utilize the HOP3D framework for generalized few-shot 3D point cloud segmentation

Machine learning researchers and engineers working on 3D point cloud segmentation tasks can benefit from this framework, as it enables adaptation to novel classes while maintaining performance on base classes

💡 Hierarchical orthogonal prototypes can effectively balance the stability-plasticity trade-off in few-shot learning

💡 HOP3D: A unified framework for generalized few-shot 3D point cloud segmentation