search

greatlog / DAN

This is an official implementation of Unfolding the Alternating Optimization for Blind Super Resolution
231 stars 41 forks source link

Wrong pth file? #17

Closed zhihongp closed 3 years ago

zhihongp commented 3 years ago

It seems pca_aniso_matrix_x4.pth and pca_matrix.pth are identical, while one is anisotropic and the other isotropic. Please check

greatlog commented 3 years ago

It seems that we have made a mistake about the pca_matrix when training DANv1 under setting 1 for scale x4. Actually, we have tested the same model for both setting 1 and setting 2. The results suggest that models trained under setting 1 are also able to generalize to setting 2.