Learning Rotation-Equivariant Features for Visual Correspondence | Read Paper on Bytez