Open andyye1999 opened 1 year ago
when you train speech super-resolution use mask method but high frequency is none mask * 0 = 0 why use mask but not mapping？
when you train speech super-resolution use mask method but high frequency is none mask * 0 = 0 why use mask but not mapping？