I'm currently working on generating images using the DWD dataset with the Style Swap technique mentioned in your paper, but the results aren't ideal. The positions of objects in the generated images are not consistent with those in the source domain images. Could you explain what might be causing this issue? Additionally, I'd like to ask if the repository provides the code for the Multi-Resolution Latent Fusion described in Section 3.4 of the paper.
I'm currently working on generating images using the DWD dataset with the Style Swap technique mentioned in your paper, but the results aren't ideal. The positions of objects in the generated images are not consistent with those in the source domain images. Could you explain what might be causing this issue? Additionally, I'd like to ask if the repository provides the code for the Multi-Resolution Latent Fusion described in Section 3.4 of the paper.