Text this: Best next-viewpoint recommendation by selecting minimum pose ambiguity for category-level object pose estimation