Essentially, for our network to upsample a point cloud, it should learn local geometry patterns from the objects.
对于上采样点云网络而言,他应该去学习局部几何模式从原有的对象当中。
我们的目标是针对点云的下游任务,做上采样。对于点云任务的上采样网络而言,他本身应该去学习local geometry patterns,所以,我们可以考虑将其做成patch-level。
local geometry patterns
patch-level