Abstract: Highlights•Develops a multi-class-one-model architecture that is both accurate and efficient.•Utilizes dual-prompts to better transfer knowledge to the 3D point cloud AD tasks.•Proposes Ano3D method for generating synthetic 3D point cloud anomaly samples•Achieving superior results across multiple evaluation metrics.
External IDs:dblp:journals/eswa/WangXCXHDT26
Loading