Attribute-Aware Face Aging With Wavelet-Based Generative Adversarial Networks

“To be specific, we employ an hourglass-shaped fully convolutional network as the generator, which has achieved success in previous image translation studies [6, 29].”

“All face images are cropped and aligned according to the five facial landmarks detected by MTCNN [27].”

“To be specific, we utilize the VGG-Face descriptor [14], denoted by φ, to extract the identity related semantic representation of a face image.”

“Unlike regular GANs [5], we adopt least square loss instead of negative log likelihood loss for that margins between generated samples and the decision boundary in the feature space are also minimized, which further improves the quality of synthesized images [12].”

“Unlike regular GANs [5], we adopt least square loss instead of negative log likelihood loss for that margins between generated samples and the decision boundary in the feature space are also minimized, which further improves the quality of synthesized images [12].”

“[28] proposed a conditional adversarial autoencoder (CAAE) to achieve age progression and regression by traversing in low-dimensional manifold.”

“been proposed to address this problem in the last two decades [8, 20, 19, 21, 7].”

“Data samples in MORPH are color images of near-frontal faces exhibiting neutral expressions under uniform and moderate illumination with simple background.”

“These proposed methods are generally divided into three groups, including physical model-based methods [21, 22], prototype-based methods [11, 23], and deep learning-based methods [9, 15, 25, 30, 32].”

“There are many researches that venture GANs face aging as image-to-image translation [19]–[23].”

“To mitigate this, some approaches introduce wavelet-based discriminators to enhance the realism of the generated images [13, 14].”

“Generative approaches [3, 23, 28, 33, 35] leverage the idea of generative and adversarial networks (GANs) [14] and concentrate on stimulating aging process to synthesize target facial images at different ages, and then minimizing the reconstruction error between real and generated images.”

“Similar techniques can be applied to face photos for testing and improving finer-grained VDL tasks like face recognition [53, 70], face aging [50, 54], crowd counting [49, 71], etc. Validation.”

“To solve the above problems, Liu et al. [43] proposed a wavelet based generative countermeasure network, which embedded the attribute vector of face image into the model generator and discriminator to encourage each synthetic face age image to be consistent with the attributes of the corresponding…”

“Regarding the face aging problem, many research were going on in the recent past from identifying the age of person from a facial image [1] to producing different age images of single person [19].”

“Recent research [20] showed that multi-level discriminators integrated with wavelet-based information decomposition can help to extract features related to identity and age for face aging.”

“For instance, wavelet-based approaches [9], [10] decompose images into multiple scales in the frequency domain to capture aging textural details.”

“Despite the extensive array of methodologies [5]–[7], [23], [24] employed in the realm of facial age editing, the task persists in its intricacy and challenges.”

“For instance, the idea of image encryption using adversarial generative networks has been discussed by Liu et al. [11].”

“Liu et al. [14] developed a grounded in generative adversarial networks (GANs).”

“In recent years, models based on GANs have been extensively utilized in the field of face aging and have made significant progress [10,11,12,13,14] .”

“With the success of deep convolution networks in image generation [13,14], Generative Adversarial Networks (GANs) based models are developed as powerful tools for age pro-123 gression [15–19].”

“In summary, our work contributions are as follows: (1) Using the Haar wavelet transform algorithm [23,24] to process the approximate information and detailed information of the image, and decompose a specific image into images of different frequencies.”

“Liu et al. [17] propose an attribute-aware facial aging model with wavelet-based GANs to address the matching ambiguity between young and aged face images inherent to the unpaired training data.”

“The recent use of wavelets in deep generative models has been mostly around GAN (Liu, Li, and Sun 2019; Angles and Mallat 2018).”

“Although some techniques [13], [14] Authorized licensed use limited to the terms of the applicable license agreement with IEEE.”