You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In an era where images and visual content dominate our digital landscape, theability to manipulate and personalize these images has become a necessity.Envision seamlessly substituting a tabby cat lounging on a sunlit window sillin a photograph with your own playful puppy, all while preserving the originalcharm and composition of the image. We present Photoswap, a novel approach thatenables this immersive image editing experience through personalized subjectswapping in existing images. Photoswap first learns the visual concept of thesubject from reference images and then swaps it into the target image usingpre-trained diffusion models in a training-free manner. We establish that awell-conceptualized visual subject can be seamlessly transferred to any imagewith appropriate self-attention and cross-attention manipulation, maintainingthe pose of the swapped subject and the overall coherence of the image.Comprehensive experiments underscore the efficacy and controllability ofPhotoswap in personalized subject swapping. Furthermore, Photoswapsignificantly outperforms baseline methods in human ratings across subjectswapping, background preservation, and overall quality, revealing its vastapplication potential, from entertainment to professional editing.
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)
The text was updated successfully, but these errors were encountered: