A Survey of Face Age Editing Based on Generative Adversarial Networks and Diffusion Models

doi:10.11871/jfdc.issn.2096-742X.2025.01.003

Abstract

Abstract:

[Purpose] In recent years, deep generative models have made significant progress in the task of facial age editing. This paper summarizes facial age editing methods based on deep generative models such as Generative Adversarial Networks (GANs) and diffusion models. [Methods] This survey first introduces the basic concepts of face age editing, relevant datasets, and evaluation metrics. It then analyzes the applications of commonly used GANs, Diffusion Models, and their variants in age editing tasks. The performance of existing models in terms of age accuracy, identity consistency, and image quality is summarized, and the suitability of different evaluation metrics is discussed. [Results] Age editing technology based on GANs and Diffusion Models have achieved significant improvements in image quality and age prediction accuracy. However, challenges remain in generating fine details, particularly when dealing with large age gaps. [Conclusions] Future research in face age editing can further enhance model generation capability and application effects by developing larger, higher-quality datasets and integrating 3D face reconstruction technology with efficient sampling algorithms from Diffusion Models.

Key words: deep learning, generative adversarial networks, attributes editing, diffusion models, face age editing

JIN Jiali, GAO Siyuan, GAO Manda, WANG Wenbin, LIU Shaozhen, SUN Zhenan. A Survey of Face Age Editing Based on Generative Adversarial Networks and Diffusion Models[J]. Frontiers of Data and Computing, 2025, 7(1): 38-55, https://cstr.cn/32002.14.jfdc.CN10-1649/TP.2025.01.003.

Figures/Tables 17

Fig.1

Fig.2

Table 1

Fig.3

Fig.4

Table 2

Fig.5

Fig.6

Table 3

Fig.7

Fig.8

Table 4

Fig.9

Table 5

Fig.10

Table 6

Fig.11

References 53

[1]	PANIS G, LANITIS A, TSAPATSOULIS N, et al. Overview of Research on Facial Aging using the FG-NET Aging Database[J]. IET Biometrics, 2016, 5(2): 37-46.
[2]	RICANEK K, TESAFAYE T. MORPH: A Longitudinal Image Database of Normal Adult Age-progression[C]. Proceedings of the IEEE International Conference on Automatic Face and Gesture Recognition, 2006: 341345.
[3]	CHEN B C, CHEN C S, HSU W H. Cross-age Reference Coding for Age-invariant Face Recognition and Retrieval[C]. Proceedings of the European Conference on Computer Vision, 2014: 768-783.
[4]	Eidinger E, Enbar R, Hassner T. Age and Gender Estimation of Unfiltered Faces[J]. IEEE Transactions on Information Forensics and Security, 2014, 9(12): 2170-2179.
[5]	ROTHE R, TIMOFTE R, VAN GOOL L. DEX: Deep EXpectation of Apparent Age from a Single Image[C]. Proceedings of the IEEE International Conference on Computer Vision Workshops, 2015: 252-257.
[6]	ZHANG Z, SONG Y, QI H. Age Progression/Regression by Conditional Adversarial Autoencoder[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 5810-5818.
[7]	OR-EL R, SENGUPTA S, FRIED O, et al. Lifespan Age Transformation Synthesis[C]. Proceedings of the European Conference on Computer Vision, 2020: 739-755.
[8]	LIU Z, LUO P, WANG X, et al. Deep Learning Face Attributes in the Wild[C]. Proceedings of the IEEE International Conference on Computer Vision, 2015: 3730-3738.
[9]	KARRAS T, TIMO A, SAMULI L, et al. Progressive Growing of GANs for Improved Quality, Stability, and Variation[C]. Proceedings of the International Conference on Learning Representations, 2018: 1-26.
[10]	Yi D, Lei Z, LI S Z. Age Estimation by Multi-Scale Convolutional Network[C]. Asian Conference on Computer Vision, 2014: 144-158.
[11]	SALIMANS T, GOODFELLOW I, ZAREMBA W, et al. Improved Techniques for Training GANs[C]. Proceedings of the Annual Conference on Neural Information Processing Systems, 2016: 2234-2242.
[12]	SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethinking the Inception Architecture for Computer Vision[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 2818-2826.
[13]	HEUSEL M, RAMSAUER H, UNTERTHINER T, et al. GANs Trained by a Two Time-scale Update Rule Converge to a Local Nash Equilibrium[C]. Proceedings of the Annual Conference on Neural Information Processing Systems, 2017: 6627-6638.
[14]	ZHANG R, ISOLA P, EFROS A A, et al. The Unreasonable Effectiveness of Deep Features as a Perceptual Metric[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 586-595.
[15]	DENG J, GUO J, XUE N, et al. Arcface: Additive Angular Margin Loss for Deep Face Recognition[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019: 4690-4699.
[16]	CAO Q, SHEN L, XIE W, et al. VGGFace2: A Dataset for Recognising Faces across Pose and Age[C]. IEEE Proceedings. Institute of Electrical and Electronics Engineers, 2018: 67-74.
[17]	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative Adversarial Networks[J]. Communications of the ACM, 2020, 63(11): 139-144.
[18]	HO J, JAIN A, ABBEEL P. Denoising Diffusion Probabilistic Models[C]. Proceedings of the 34th International Conference on Neural Information Processing Systems, 2020: 6840-6851.
[19]	SONG J, MENG C, ERMON S. Denoising Diffusion Implicit Models[C]. International Conference on Learning Representations, 2021: 1-22.
[20]	YANG H, HUANG D, WANG Y, et al. Learning Face Age Progression: A Pyramid Architecture of GANs[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 31-39.
[21]	LIU Y, LI Q, SUN Z. Attribute-Aware Face Aging with Wavelet-Based Generative Adversarial Networks[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019: 11877-11886.
[22]	SONG J, ZHANG J, GAO L, et al. Dual Conditional GANs for Face Aging and Rejuvenation[C]. International Joint Conference on Artificial Intelligence, 2018: 899-905.
[23]	SONG J, ZHANG J, GAO L, et al. AgeGAN++: Face Aging and Rejuvenation with Dual Conditional GANs[J]. IEEE Transactions on Multimedia, 2022, 24(2): 791-804.
[24]	YANG H, HUANG D, WANG Y, et al. Learning Continuous Face Age Progression: A Pyramid of GANs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 43(2): 499-515.
[25]	HELJAKKA A, SOLIN A, KANNALA J. Recursive Chaining of Reversible Image-to-image Translators for Face Aging[C]. Advanced Concepts for Intelligent Vision Systems, 2018: 309-320.
[26]	HUANG Z, CHEN S, ZHANG J, et al. PFA-GAN: Progressive Face Aging with Generative Adversarial Network[J]. IEEE Transactions on Information Forensics and Security, 2021, 16(4): 2031-2045.
[27]	Antipov G, Baccouche M, Dugelay J L. Face Aging with Conditional Generative Adversarial Networks[C]. IEEE International Conference on Image Processing, 2017: 2089-2093.
[28]	Wang Z, Tang X, Luo W, et al. Face Aging with Identity-Preserved Conditional Generative Adversarial Networks[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7939-7947.
[29]	LI P, HUANG H, HU Y, et al. Hierarchical Face Aging through Disentangled Latent Characteristics[C]. Proceedings of the European Conference on Computer Vision, 2020: 86-101.
[30]	ZHAO J, CHENG Y, CHENG Y, et al. Look across Elapse: Disentangled Representation Learning and Photorealistic Cross-age Face Synthesis for Age-invariant Face Recognition[C]. Proceedings of the AAAI conference on Artificial Intelligence, 2019: 9251-9258.
[31]	JEON S, LEE P, HONG K, et al. Continuous Face Aging Generative Adversarial Networks[C]. Proceedings of the IEEE International Conference on Acoustics, 2021: 1995-1999.
[32]	MAKHMUDKHUJAEV F, HONG S, PARK I K. Re-aging Gan: Toward Personalized Face Age Transformation[C]. Proceedings of the IEEE International Conference on Computer Vision, 2021: 3908-3917.
[33]	GOMEZ-TRENADO G, LATHUILIÈRE S, MESEJO P, et al. Custom Structure Preservation in Face Aging[C]. European Conference on Computer Vision, 2022: 565-580.
[34]	BAHDANAU D, CHO K H, BENGIO Y. Neural Machine Translation by Jointly Learning to Align and Translate[C]. International Conference on Learning Representations, 2015: 1-15.
[35]	LI Q, LIU Y, SUN Z. Age progression and regression with spatial attention modules[C]. Proceedings of the AAAI conference on artificial intelligence, 2020: 11378-11385.
[36]	KWAK J, HAN D K, KO H. CAFE-GAN: Arbitrary Face Attribute Editing with Complementary Attention Feature[C]. European Conference on Computer Vision, 2020: 524-540.
[37]	ZHU H, HUANG Z, SHAN H, et al. Look Globally, Age Locally: Face Aging with an Attention Mechanism[C]. International Conference on Acoustics, Speech and Signal Processing, 2020: 1963-1967.
[38]	SHI C, ZHANG J, YAO Y, et al. CAN-GAN: Conditioned-attention Normalized GAN for Face Age Synthesis[J]. Pattern Recognition Letters, 2020, 138(2): 520-526.
[39]	LIU Y, LI Q, SUN Z, et al. A3GAN: an Attribute-aware Attentive Generative Adversarial Network for Face Aging[J]. IEEE Transactions on Information Forensics and Security, 2021, 16(2): 2776-2790.
[40]	CHANDALIYA P K, NAIN N. AW-GAN: Face Aging and Rejuvenation Using Attention with Wavelet GAN[J]. Neural Computing and Applications, 2023, 35(3): 2811-2825.
[41]	KARRAS T, LAINE S, AILA T. A Style-based Generator Architecture for Generative Adversarial Networks[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019: 4401-4410.
[42]	HE S, LIAO W, YANG M Y, et al. Disentangled Lifespan Face Synthesis[C]. Proceedings of the IEEE International Conference on Computer Vision, 2021: 3877-3886.
[43]	ALALUF Y, PATASHNIK O, COHEN-OR D. Only a Matter of Style: Age Transformation Using a Style-based Regression Model[J]. ACM Transactions on Graphics, 2021, 40(4): 1-12.
[44]	LIU L, YU H, WANG S, et al. Learning Shape and Texture Progression for Young Child Face Aging[J]. Signal Processing: Image Communication, 2021, 93(116): 1-10.
[45]	MA W, ZHOU Y, HE J. Semi-supervised Face Aging and Rejuvenating[J]. Journal of Electronic Imaging, 2021, 30(2): 023003-023003.
[46]	RADFORD A, KIM J W, HALLACY C, et al. Learning Transferable Visual Models from Natural Language Supervision[C]. International Conference on Machine Learning, 2021: 8748-8763.
[47]	CHEN X, LATHUILIÈRE S. Face Aging via Diffusion-based Editing[C]. British Machine Vision Conference, 2023: 1-13.
[48]	LI P, WANG R, HUANG H, et al. Pluralistic Aging Diffusion Autoencoder[C]. Proceedings of the IEEE International Conference on Computer Vision, 2023: 22613-22623.
[49]	Banerjee S, Mittal G, Joshi A, et al. Identity-Aware Facial Age Editing Using Latent Diffusion[J]. IEEE Transactions on Biometrics, Behavior, and Identity Science, 2024, 6(4): 443-457
[50]	MOSCHOGLOU S, PAPAIOANNOU A, SAGONAS C, et al. Agedb: the First Manually Collected, in-the-wild Age Database[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2017: 51-59.
[51]	ROMBACH R, BLATTMANN A, LORENZ D, et al. High-resolution Image Synthesis with Latent Diffusion Models[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022: 10684-10695.
[52]	Hou C, Wei G, Chen Z. High-Fidelity Diffusion-Based Image Editing[C]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024: 2184-2192.
[53]	Parihar R, Sachidanand V S, Mani S, et al. PreciseControl: Enhancing Text-to-Image Diffusion Models with Fine-Grained Attribute Control[C]. European Conference on Computer Vision, 2025: 469-487.

数据集名称	发布时间	图像数量	人物数量	年龄范围	采集环境	标签类型	人物身份
FG-Net^[1]	2002年	1002	82	0-69	非受控	年龄值	普通人
MORPH Album1^[2]	2006年	1690	515	15-68	受控	年龄值	普通人
MORPH Album2^[2]	2006年	55134	13618	16-77	受控	年龄值	普通人
CACD^[3]	2013年	163446	2000	16-62	非受控	年龄值	名人
Adience^[4]	2014年	26580	2284	0-60+	非受控	年龄段	普通人
IMDB-WIKI^[5]	2015年	523051	20284	0-100	非受控	年龄值	名人
UTKFace^[6]	2017年	20000	-	0-116	非受控	年龄值	普通人
FFHQ-Aging^[7]	2020年	70000	-	0-70+	非受控	年龄段	普通人

方法	发布时间	数据集	创新点
PA-GAN^[20]	2018年	FG-Net, MORPH, CACD	引入金字塔结构的判别器，通过多尺度方式评估生成图像的年龄特征
AW-GAN^[21]	2019年	MORPH, CACD	引入属性感知嵌入和小波包变换，增强局部纹理特征
Age GAN^[22]	2018年	UTKFace	使用Dual cGANs架构实现年龄变换
AgeGAN++^[23]	2021年	FG-Net, MORPH, CACD	引入表征解耦模块，增强身份特征与年龄特征的独立性
PA-GAN++^[24]	2021年	FG-Net, MORPH, CACD	采用多个并行判别器，专注于不同年龄组的特征学习

方法	发布时间	数据集	创新点
Heljakka等^[25]	2018年	CACD	引入了一种可逆GAN转换链，链中的每个单元都可以处理相邻年龄域之间的转换
PFA-GAN^[26]	2021年	MORPH, CACD	通过多个子网络分阶段处理人脸老化，每个子网络只负责相邻年龄组之间的老化效果

方法	发布时间	数据集	创新点
Age-GAN^[27]	2017年	IMDB-WIKI	通过更改GAN的条件输入实现面部的年龄合成
IPCGANs^[28]	2018年	CACD	引入感知损失，通过高层特征保持生成图像与输入图像在身份上的一致性
CAAE^[6]	2017年	FG-Net, MORPH, CACD	学习面部图像的高维流形和潜在空间映射
DAAE^[29]	2020年	CACD, MORPH, UTKFace, FG-NET	引入两种先验分布和身份知识蒸馏，确保特征解耦的准确性
AIM^[30]	2019年	FG-Net, MORPH, CACD	提取鲁棒的年龄无关身份特征表示
CFA-GAN^[31]	2018年	MORPH	使用正交分解的编码器特征，确保身份和年龄特征独立性
Re-Aging GAN^[32]	2021年	FG-Net, MORPH, CACD	引入年龄调制模块和自监督机制
CUSP^[33]	2023年	FFHQ	引入掩码机制允许用户自定义结构保持级别，结合风格和内容编码器生成个性化的年龄编辑图像
Li等^[35]	2020年	MORPH,CACD,UTKface	使用两个生成器分别处理年龄进展和回归任务，引入空间注意力机制，将图像的修改集中在面部年龄特征区域
CAFE-GAN^[36]	2020年	CelebA	提出了一种补充注意力特征方法，设计用于只编辑与目标属性相关的面部区域
AcGAN^[37]	2020年	MORPH	引入了注意力机制，使生成器只关注面部与年龄相关的区域
CAN-GAN^[38]	2020年	MORPH, CACD,FG-NET	分配权重来度量每个面部特征在年龄分类中的重要性，从而提高分类器区分年龄的精度
A³-GAN^[39]	2021年	FG-Net, MORPH, CACD, CelebA	将面部属性嵌入到生成器和判别器中，模型能够生成符合输入属性的老化效果，引入注意力机制修改与年龄相关的区域
AW-GAN^[40]	2022年	CACD, FG-Net,	引入了小波变换与轻量化的卷积注意力模块 (mCBAM)

方法	发布时间	数据集	创新点
LATS^[7]	2020年	FFHQ-aging	创建多个年龄锚点类别，实现了一个近似连续的年龄潜在空间
DLFS^[42]	2021年	FFHQ-aging	将面部的形状、纹理和身份特征解耦，以便在年龄转换时分别控制形状和纹理变化
SAM^[43]	2021年	FFHQ-aging	将图像编码为潜在空间中的风格向量，并应用年龄回归网络来引导编码器生成对应于目标年龄的潜在编码
Liu等^[44]	2021年	UTKFace,YAFD^[44]	引入 GD-GAN 和 TV-GAN，分别处理面部形状的几何变化和纹理变化
Ma等^[45]	2021年	UTKFace,CACD	结合 cVAE 和 cGAN 的架构，无需大量配对样本