The MA in Museum Studies and Curatorial Practices has been designed to prepare graduates for professional positions in the highly complex and diverse museum landscape in South East Asia.

外观、几何、语义三类信息通常由独立模块处理，导致模态间语义错位。现有视频扩散模型仅能生成单一模态，当需要同步生成法线和语义时，会出现物体表面法线与颜色不匹配等问题，在ScanNet测试中，传统方法的法线与RGB一致性误差平均达27.3°。