With the improvement of face recognition precision, face recognition system is used in many fields. However, the face recognition system sometimes cannot recognize the makeup face. In this paper, a new image-to-image translation algorithm based on GAN and dual learning is proposed to remove the makeup. Especially, the proposed algorithm is weakly supervised and it combines the paired and unpaired image-to-image translation model. The dual model is firstly trained using a small number of paired data, then the performance of the model is improved by large number of unpaired data. The proposed weakly-supervised image-to-image translation algorithm is applied into makeup-removal task, and the experimental results demonstrate its higher performance than other algorithms.
CITATION STYLE
Hou, X., Li, Y., & Li, T. (2017). Weakly-Supervised Dual Generative Adversarial Networks for Makeup-Removal. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10635 LNCS, pp. 603–611). Springer Verlag. https://doi.org/10.1007/978-3-319-70096-0_62
Mendeley helps you to discover research relevant for your work.