Abstract—Facial beauty analysis is a crucial subject in human culture among researchers through different applications. Recent studies used multidisciplinary approaches to examine the relation between facial traits, age, emotions, and other factors. However, facial beauty prediction is a significant visual recognition challenge for the evaluation of facial attractiveness for human perception, which requires a considerable effort due to the field’s novelty and lack of resources with a small database for facial beauty prediction. In this vein, a deep learning method has recently demonstrated remarkable qualities in facial beauty prediction. Additionally, vision Transformers have recently been introduced as novel Deep Learning approaches and have presented a strong performance in a number of applications. The key issue is that vision transformer performs significantly worse than ResNet when trained on a small ImageNet database. In this paper, we propose to tackle the difficulties of facial beauty prediction, using vision transformers as opposed to feature extraction based on Convolutional Neural Networks commonly used in traditional methods. Moreover, we define and optimize a set of hyper-parameters according to the SCUT-FBP5500 benchmark dataset the model obtains 0.9534 Pearson Coefficient. Experimental results indicated that using this proposed network leads to better predicting of facial beauty closer to human evaluation than conventional technology that provides facial beauty assessment.
 

Index Terms—Facial beauty prediction, vision transformer, deep learning, convolutional neural networks, performance evaluation

Cite: Djamel Eddine Boukhari, Ali Chemsa, and Riadh Ajgou, "Facial Beauty Prediction Based on Vision Transformer," International Journal of Electrical and Electronic Engineering & Telecommunications