In this paper, we propose a novel GAN inversion approach to semantic manipulation of out-of-range images that are geometrically unaligned with the training images of a GAN model. To find a latent code that is semantically editable, our approach invert...
In this paper, we propose a novel GAN inversion approach to semantic manipulation of out-of-range images that are geometrically unaligned with the training images of a GAN model. To find a latent code that is semantically editable, our approach inverts an input out-of-range image into an alternative latent space than the original latent space. We also propose a regularized inversion method to find a proper solution that supports semantic manipulation in the alternative space. Our experiments show that our approach effectively supports semantic manipulation of out-of-range images with geometric transformations.