An Effective Hit-or-Miss Layer Favoring Feature Interpretation as Learned Prototypes Deformations | Read Paper on Bytez