confused

[JinglUO703]

In rotationnet and view GCN and the two works, the modelnet40, 20view used are obtained after the 3D object is rotated. Why do you put the result of rotation in aligned in your paper?

[weixmath]

The results reported in rotationnet and view-GCN correspond to the 'aligned' version in this paper. Shapes are aligned before rendering. When shapes are rotated, their performance drop dramatically. You can refer to 'Equivariant multi-view network', which reported the same results.