我們提出了用於人臉年齡轉換的Cycle Consistency and Conditional Projection GAN (CCCP-GAN)。提出的 CCCP-GAN 由編碼器、解碼器和判別器組成，其中編碼器將帶有目標年齡組標籤的輸入圖像轉換為身份潛在編碼和年齡潛在編碼，並將最新合成網絡StyleGAN2的解碼器進行修改，將身份潛在代碼和年齡潛在代碼作為輸入，生成保留輸入身份並處於目標年齡的人臉。判別器由提出的條件多任務投影 (CMP) 組成，以更好地在生成的圖像上生成所需的年齡特徵，從而在單一模型中實現年齡進展和回歸。最後，我們利用三元組損失和循環損失來保留輸入身份並在輸出處獲得所需的年齡分佈。在本論文的新穎之處包括： 1. 將年齡特徵與所需年齡組標籤融合的 CMP 判別器，能夠有效監督單個生成器執行年齡進展和回歸； 2. 對最新風格轉移網路施加循環一致性，以在單一框架內實現保持身分的臉部年齡轉換； 3. 與最先進的方法相比，生成具有更好性能的高分辨率年齡轉移圖像； 4. 提出了一種新的協議，用於使用新的指標和基準來評估跨大年齡差距的年齡轉變。

We propose the Cycle Consistency and Conditional Projection GAN for facial age transformation. The CCCP-GAN is composed of an encoder, a decoder and a discriminator. The encoder converts an input image with a target age group label to an identity latent code and an age latent code. The decoder, developed on the StyleGAN2 synthesis network, takes the identity latent code and the age latent code as inputs to generate a face which preserves the input identity and is in the target age group. The discriminator is made of the proposed Conditional Multitask Projection (CMP) to better generate the desired age traits on the generated images, enabling both age progression and regression in a unified model. We exploit the triplet loss and cycle loss to preserve the input identity and attain the desired age distribution at the output. The novelties of this study include the following: 1. The CMP discriminator that fuses age features with the desired age group labels able to effectively supervise a single generator to perform age progression and regression; 2. A cycle consistency imposed on a latest style transfer network for achieving identity-preserving facial age transformation in a unified framework; 3. Generation of high-resolution age-transferred images with a better performance compared to state-of-the-art approaches; 4. A new protocol proposed for evaluating the age transformation across large age gaps with a novel metrics and benchmarks.

[1] Zhang, Zhifei, Yang Song, and Hairong Qi. "Age progression/regression by conditional adversarial autoencoder." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[2] Li, Peipei, et al. "Global and local consistent age generative adversarial networks." 2018 24th International Conference on Pattern Recognition (ICPR). IEEE, 2018.
[3] Yang, Hongyu, et al. "Learning face age progression: A pyramid architecture of gans." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
[4] Liu, Yunfan, Qi Li, and Zhenan Sun. "Attribute-aware face aging with wavelet-based generative adversarial networks." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
[5] Fu, Yun, Guodong Guo, and Thomas S. Huang. "Age synthesis and estimation via faces: A survey." IEEE transactions on pattern analysis and machine intelligence 32.11 (2010): 1955-1976.
[6] Shen, Yujun, et al. "Interpreting the latent space of gans for semantic face editing." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020.
[7] Yang, Hongyu, et al. "Learning continuous face age progression: A pyramid of gans." IEEE transactions on pattern analysis and machine intelligence (2019).
[8] Karras, Tero, et al. "Analyzing and improving the image quality of stylegan." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020.
[9] J.-Y. Zhu, T. Park, P. Isola, and A. A. Efros, “Unpaired image-to-image translation using cycle-consistent adversarial networks,” in ICCV, 2017
[10] F. Schroff, D. Kalenichenko, and J. Philbin, “Facenet: A unified embedding for face recognition and clustering,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 815–823
[11] Huang, Xun, and Serge Belongie. "Arbitrary style transfer in real-time with adaptive instance normalization." Proceedings of the IEEE International Conference on Computer Vision. 2017.
[12] Karras, Tero, Samuli Laine, and Timo Aila. "A style-based generator architecture for generative adversarial networks." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
[13] A. Karnewar and O. Wang, “Msg-gan: Multi-scale gradients for generative adversarial networks,” in CVPR, 2020.
[14] E. L. Denton, S. Chintala, R. Fergus et al., “Deep generative image models using a laplacian pyramid of adversarial networks,” in NIPS, 2015
[15] A. Odena, C. Olah, and J. Shlens, “Conditional image synthesis with auxiliary classifier gans,” in ICML, 2017.
[16] S. Reed, Z. Akata, X. Yan, L. Logeswaran, B. Schiele, and H. Lee, “Generative adversarial text to image synthesis,” arXiv preprint arXiv:1605.05396, 2016
[17] H. Zhang, T. Xu, H. Li, S. Zhang, X. Wang, X. Huang, and D. N. Metaxas, “Stackgan: Text to photo-realistic image synthesis with stacked generative adversarial networks,” in ICCV, 2017.
[18] T. Kim, M. Cha, H. Kim, J. K. Lee, and J. Kim, “Learning to discover cross-domain relations with generative adversarial networks,” arXiv preprint arXiv:1703.05192, 2017.
[19] M. Mirza and S. Osindero, “Conditional generative adversarial nets,” arXiv preprint arXiv:1411.1784, 2014.
[20] V. Dumoulin, I. Belghazi, B. Poole, O. Mastropietro, A. Lamb, M. Arjovsky, and A. Courville, “Adversarially learned inference,” arXiv preprint arXiv:1606.00704, 2016.
[21] K. Sricharan, R. Bala, M. Shreve, H. Ding, K. Saketh, and J. Sun, “Semi-supervised conditional gans,” arXiv preprint arXiv:1708.05789, 2017.
[22] T. Miyato and M. Koyama, “cgans with projection discriminator,” arXiv preprint arXiv:1802.05637, 2018.
[23] Z. Wang, X. Tang, W. Luo, and S. Gao, “Face aging with identity-preserved conditional generative adversarial networks,” in CVPR, 2018.
[24] R. Or-El, S. Sengupta, O. Fried, E. Shechtman, and I. Kemelmacher-Shlizerman, “Lifespan age transformation synthesis,” in ECCV, 2020.
[25] K. Ricanek and T. Tesafaye, “Morph: A longitudinal image database of normal adult age-progression,” in FG, 2006.
[26] B.-C. Chen, C.-S. Chen, and W. H. Hsu, “Face recognition and retrieval using cross-age reference coding with cross-age celebrity dataset,” TMM, 2015.
[27] M. Inc., “Face++ research toolkit,” http://www.faceplusplus. com.
[28] H. Pan, H. Han, S. Shan, and X. Chen, “Mean-variance loss for deep age estimation from a face,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 5285–5294.
[29] A. Bulat and G. Tzimiropoulos, “How far are we from solving the 2d & 3d face alignment problem? (and a dataset of 230,000 3d facial landmarks),” in ICCV, 2017.
[30] Z. He, M. Kan, S. Shan, and X. Chen, “S2gan: Share aging factors across ages and share aging trends among individuals,” in ICCV, 2019.
[31] L. Tran, X. Yin, and X. Liu, “Disentangled representation learning gan for pose-invariant face recognition,” in CVPR, 2017.
[32] J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in CVPR, 2019.
[33] Karras, Tero, et al. "Progressive growing of gans for improved quality, stability, and variation." arXiv preprint arXiv:1710.10196 (2017).
[34] LeCun, Yann, Corinna Cortes, and C. J. Burges. "MNIST handwritten digit database." (2010): 18.
[35] Panis, Gabriel, et al. "Overview of research on facial ageing using the FG-NET ageing database." Iet Biometrics 5.2 (2016): 37-46.