Open TomSuen opened 3 months ago
Thank you for your interest in our work! To clarify, it is not a copy. For each operation, the latent patch at the initial position is used as the optimization target for different timesteps. The optimization loss function is defined by Eq. 8 in our paper. Additionally, Fig. 5 illustrates that if the optimization target patch is not fixed, cumulative errors can lead to drag operation deviation and image collapse.
Hi, such a wonderful work! I want to know when a drag operation is performed on a latent noised image, is the value of that position directly copied to the target position?