MorphGANFormer

This is the offcial implementation of paper 'MorphGANFormer: Transformer-based Face Morphing and De-Morphing'. Code will be coming soon.

Link:

[PDF] [Arxiv]

This repository contains the implementation of:

• MorphGANFormer
• Morphing and De-morphing

Our model is based on the paper: GANformer: Generative Adversarial Transformers [Github] [Arxiv]

Introduction

Inspired by GANformer, we introduce a transformer-based face morphing algorithm. Special loss functions are designed to support the optimization of face morphing process. We extend the study of transformer-based face morphing to demorphing by presenting an effective defense strategy with access to a reference image using the same generator of MorphGANFormer. Such demorphing is conceptually similar to unmixing of hyperspectral images but operates in the latent (instead of pixel) space.

Environment

• Python 3.6 or 3.7 are supported.
• Pytorch >= 1.8.
• CUDA 10.0 toolkit and cuDNN 7.5.
• See requirements.txt for the required python packages and run pip install -r requirements.txt to install them.

Face Morphing Pipeline

Face De-Morphing Pipeline

Face Latent code Optimization