AttGAN based simultaneous modification of the given multiple face attributes

Facial attributes are interesting because they show the identity of a person, and manipulating these facial attributes is challenging as it requires the preservation of the identity while making the required modifications to the appearance. Many deep-neural-network-basedof approaches have been proposed for attribute manipulation. However, these existing approaches are incapable effectively modifying the attributes simultaneously. In this paper, we address the problem of simultaneous modification of face attributes by manipulating the class information of the attributes using an attribute generative adversarial network. Based on the proposed approach, the latent representation of the given image and its corresponding attribute information are used to modify the appearance by negating the given attribute information. Experiments on the CelebA dataset show that our method effectively performs simultaneous attribute editing with the preservation of other intact facial details.