BiTrackGAN: Cascaded CycleGANs to Constraint Face Aging

Tsung-Han Kuo,Zhenge Jia,Tei-Wei Kuo,Jingtong Hu

2023-04-22

Abstract:With the increased accuracy of modern computer vision technology, many access control systems are equipped with face recognition functions for faster identification. In order to maintain high recognition accuracy, it is necessary to keep the face database up-to-date. However, it is impractical to collect the latest facial picture of the system's user through human effort. Thus, we propose a bottom-up training method for our proposed network to address this challenge. Essentially, our proposed network is a translation pipeline that cascades two CycleGAN blocks (a widely used unpaired image-to-image translation generative adversarial network) called BiTrackGAN. By bottom-up training, it induces an ideal intermediate state between these two CycleGAN blocks, namely the constraint mechanism. Experimental results show that BiTrackGAN achieves more reasonable and diverse cross-age facial synthesis than other CycleGAN-related methods. As far as we know, it is a novel and effective constraint mechanism for more reason and accurate aging synthesis through the CycleGAN approach.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems Addressed by the Paper The paper primarily addresses the issue in facial recognition systems where the database needs constant updating due to the user's facial changes over time. Specifically, the paper proposes a new method called **BiTrackGAN**, aimed at improving the synthesis of cross-age facial images through the following approaches: 1. **Constraint Mechanism**: - A new constraint mechanism is proposed to generate more reasonable and smooth cross-age facial images within the CycleGAN method. 2. **Underlying Training Method**: - An underlying training method is proposed to train a translation pipeline composed of two cascaded CycleGAN blocks (BiTrackGAN), thereby inducing an ideal intermediate state as a constraint mechanism. 3. **Enhancing Synthesis Diversity**: - The underlying training method enhances the diversity and reasonableness of the synthesized images, addressing the issue of mode collapse. With these improvements, BiTrackGAN can better maintain the update of facial databases in practical applications and improve the accuracy of facial recognition systems. Particularly in application scenarios requiring cross-age image synthesis, such as tumor growth prediction in medical images, BiTrackGAN has great potential.

BiTrackGAN: Cascaded CycleGANs to Constraint Face Aging

Cycle Face Aging Generative Adversarial Networks

Face Aging with Contextual Generative Adversarial Nets

Age-Oriented Face Synthesis with Conditional Discriminator Pool and Adversarial Triplet Loss

Learning Continuous Face Age Progression: A Pyramid of GANs

Attention-aware conditional generative adversarial networks for facial age synthesis

A Unified Framework for Biphasic Facial Age Translation With Noisy-Semantic Guided Generative Adversarial Networks

Face aging with conditional generative adversarial networks

Continuous Face Aging Generative Adversarial Networks

Global and Local Consistent Age Generative Adversarial Networks

Look Globally, Age Locally: Face Aging With an Attention Mechanism

Learning Face Age Progression: A Pyramid Architecture of GANs

Continuous Face Aging via Self-estimated Residual Age Embedding

A 3 GAN: An Attribute-Aware Attentive Generative Adversarial Network for Face Aging

Two Birds with One Stone: Transforming and Generating Facial Images with Iterative GAN

Age Progression and Regression with Spatial Attention Modules

Two Birds with One Stone: Iteratively Learn Facial Attributes with GANs.

AgeGAN++: Face Aging and Rejuvenation with Dual Conditional GANs.

Building Face Ageing Model Using Face Synthesis

AgeFlow: Conditional Age Progression and Regression with Normalizing Flows

Refining CycleGAN with attention mechanisms and age-Aware training for realistic Deepfakes