SumajGAN: Supervised Transfer of Face Makeup with Deep Generative Adversarial Networks
DOI:
https://doi.org/10.26439/ciis2020.5503Keywords:
autoencoder, CNN, BeautyGAN, PairedCycleGAN, BeautyGlowAbstract
The challenge of transferring makeup from one image to another is already solved by BeautyGAN, PairedCycleGAN and Beauty Glow. These models managed to meet the aforementioned challenge using a semi-supervised learning approach which solves the problem of obtaining an aligned makeup dataset but at the expense of a high computing power. Therefore, in this research, an aligned image dataset was developed and, additionally, a makeup transfer model was proposed using a supervised approach. The dataset consisted of 5,400 groups of images: each group of images was composed of a person’s no-makeup face image, a reference makeup face image, and a reference makeup face image transferred into the person’s no-makeup face image. The model proposed in this research is called SumajGAN. It is made of a PatchGAN-type discriminator and a two-input generator based on an autoenco der. Several experiments were conducted, and the best result achieved a mean absolute error of 0.021658644 and a high-resolution makeup transfer. The SumajGAN model has mana ged to achieve the objective by reducing the training time of models such as BeautyGAN, PairedCycleGAN and Beauty Glow.
Downloads
References
Chen, H., Hui, K., Wang, S., Tsao, L., Shuai, H., y Cheng, W. (2019). BeautyGlow: On-Demand Makeup Transfer Framework with Reversible Generative Network. IEEE Xplore. https://bit.ly/2wAPZDk
Chang, H., Lu, J., Yu, F., y Finkelstein, A. (2018) PairedCycleGAN: Asymmetric Style Transfer for Applying and Removing Makeup [Sesión de conferencia]. 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). doi 10.1109/CVPR.2018.00012
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A.,y Bengio, Y. (2020). Generative Adversarial Networks. Commun. ACM, 63(11), 139-144. https://doi.org/10.1145/3422622
Gulli, A., y Pal, S. (2017). Deep Learning with KerasImplement Neural Networks with Keras on Theano and TensorFlow. Packt Publishing.
Guo, D., y Sim, T. (2009). Digital face makeup by example. 2009 IEEE Conference on Computer Vision and Pattern Recognition, 73-79. https://doi.org/10.1109/CVPR.2009.5206833
Isola, P., Zhu J., Zhou T., y Efros, A. (2017). Image-to-Image Translation with Conditional Adversarial Networks. 2017 [Sesión de conferencia]. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (pp. 5967-5976). 10.1109/CVPR.2017.632.
Kingma, D., y Dhariwal, P. (2018). Glow: Generative Flow with Invertible 1×1 Convolutions. arXiv:1807.03039v2.
Li, T., Qian, R., Dong, C., Liu, S., y Yan, Q. (2016). Makeup Like a Superstar: Deep Localized Makeup Transfer Network [Sesión de conferencia]. IJCAI’16 Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence), (pp. 2568-2575).
Li, T., Qian, R., Dong, C., Liu, S., y Yan, Q. (2018). BeautyGAN: Instance-level Facial Makeup Transfer with Deep Generative Adversarial Network [Sesión de conferencia]. 2018 ACM Multimedia Conference on Multimedia Conference (ACMMM), (pp. 645-653). 10.1145/3240508.3240618.
Zhu, J., Park, T., Isola, P., y Efros, A. (2017). Unpaired image-to-image translation using cycle-consistent adversarial networks. ArXiv preprint arXiv:1703.10593
