GAN Inversion for Image Editing via Unsupervised Domain Adaptation
This addresses a practical limitation in image editing for users dealing with common low-quality inputs, though it is incremental as it builds on existing GAN inversion techniques.
The paper tackles the problem of GAN inversion struggling with low-quality images by proposing an unsupervised domain adaptation method, achieving a PSNR of 22.14 on the FFHQ dataset and performing comparably to supervised approaches.
Existing GAN inversion methods work brilliantly in reconstructing high-quality (HQ) images while struggling with more common low-quality (LQ) inputs in practical application. To address this issue, we propose Unsupervised Domain Adaptation (UDA) in the inversion process, namely UDA-inversion, for effective inversion and editing of both HQ and LQ images. Regarding unpaired HQ images as the source domain and LQ images as the unlabeled target domain, we introduce a theoretical guarantee: loss value in the target domain is upper-bounded by loss in the source domain and a novel discrepancy function measuring the difference between two domains. Following that, we can only minimize this upper bound to obtain accurate latent codes for HQ and LQ images. Thus, constructive representations of HQ images can be spontaneously learned and transformed into LQ images without supervision. UDA-Inversion achieves a better PSNR of 22.14 on FFHQ dataset and performs comparably to supervised methods.