Comment by diyer22
It's just a coincidence—the guided images used for ZSCG all come from Celeb-A, whereas the DDN model was trained only on FFHQ.
Besides, I feel the red shoulder strap/blob is reconstructed rather poorly.
It's just a coincidence—the guided images used for ZSCG all come from Celeb-A, whereas the DDN model was trained only on FFHQ.
Besides, I feel the red shoulder strap/blob is reconstructed rather poorly.