Beholder-Gan: Generation and Beautification of Facial Images with Conditioning on Their Beauty Level

Abstract

Beauty is in the eye of the beholder." This maxim, emphasizing the subjectivity of the perception of beauty, has enjoyed a wide consensus since ancient times. In the digital era, data-driven methods have been shown to be able to predict human-assigned beauty scores for facial images. In this work, we augment this ability and train a generative model that generates faces conditioned on a requested beauty score. In addition, we show how this trained generator can be used to "beautify" an input face image. By doing so, we achieve an unsupervised beautification model, in the sense that it relies on no ground truth target images. Our implementation is available on: https://github.com/beholdergan/Beholder-GAN.

Facial reshaping operator for controllable face beautification

“…We generate these results using the popular Helen facial image dataset (Le et al, 2012) We use the red boxes to highlight the overly modified characteristic facial parts that some end-users may want to preserve during beautification. (Diamant et al, 2019). Right: the beautified images generated by our method.…”

“…Comparisons with Semi-supervised Methods. Here, we compare our method with that of (Diamant et al, 2019), which is the most recent method that uses facial attractiveness annotations to train a beauty-conditional generative adversarial network (GAN) for semi-supervised face beautification. This method represents the state-of-the-art research on leveraging deep learning and generative modelling for face beautification.…”

“…11. It can be clearly seen that the method of (Diamant et al, 2019) modifies mainly the facial textures, resulting in smoother skins (the top two faces on the middle column) and a new facial identity (the bottom face on the middle column). In contrast, our method solely enhances the facial shapes (e.g., the thinner jawline as shown on the bottom right), without altering the original textures or complexions.…”

Body LayARs: A Toolkit for Body-Based Augmented Reality

“…An example of this are generated artificial hair colors [38] and makeup [19]. A common goal behind such methods is beautification [9,27].…”

Impact and Detection of Facial Beautification in Face Recognition: An Overview

“…Such transfer is motivated by the demand of users attempting to copy makeup styles of other individuals such as celebrities. In addition, GANs were conditioned on beauty scores [60], in order to generate realistic facial images using Progressive Growing of GANs (PGGAN) [61]. Similarly, Liu et al [62] proposed a two-stage deep network for beautification, where a multi-label CNN evaluated the quality of faces, followed by a Bayesian GANs framework, automatically generating photo-realistic beautified faces.…”