研究者们通过设计一种能提供可微分 BSP-tree 表征的网络架构来完成这个任务。由于这种表征通过隐函数编码几何图形,所以它们很容易训练。
The network is trained to reconstruct a shape using a set of convexes obtained from a BSP-tree built
The network is trained to reconstruct a shape using a set of convexes obtained from a BSP-tree built