Applying GANs for Generating Image with Varied Facial Attributes from Sketch. (August 2020)

• Record Type: Journal Article
• Authors: Phusomsai, W; Limpiyakorn, Y
• Is Part Of: Journal of physics. Volume 1619(2020)
• Publication Date: 2020-08
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1619/1/012013
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: The rapid development of GANs and its variants has shown remarkable progress in synthesizing realistic images. Image-to-image translation becomes a potent research topic due to its wide application. The translation of face sketch to color images greatly contributes to digital image processing industry. It could also help confirm the identity of the suspect or lost persons. In this paper, a model is implemented by applying GANs for generating an image with varied facial attributes from sketch. The architecture consists of two separate GANs, implemented with Pix2Pix and StarGan2 as bases. The output realistic images can be generated with manipulating a single facial attribute including, wavy hair, straight hair, and wearing glasses.