Lifespan Age Synthesis with Geometry Guidance and Decorrelation Constraint

Project Page | Paper

-¹ , -¹, -², -³, -¹
¹-, ²-, ³-

Overview

This code is the official PyTorch implementation of the paper:

Lifespan Age Synthesis with Geometry Guidance and Decorrelation Constraint
-, -, -, -, -(aut
(journal)
(arxiv link)

Preparation

Please follow this github to prepare the environments.

Training and Testing (link to the pretrained models in the colab)

Download the dataset from male and female and put them to ./datasets/males and ./datasets/females.
Training (please modify --dataroot, --name):

sh train_GGDC.sh

Testing (please modify --dataroot, --name, --which_epoch, and --checkpoing_dir):

sh test_GGDC.sh

Quick Demo

You can run the demo locally or explore it in Colab

Training/Testing on New Datasets

If you wish to train the model on a new dataset, arrange it in the following structure:

├── dataset_name                                                                                                                                                                                                       
│   ├── train<class1> 
|   |   └── image1.png
|   |   └── image2.png
|   |   └── ...                                                                                                
│   │   ├── parsings
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ...         
│   │   ├── landmarks
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ...                                                                                                                          
...
│   ├── train<classN> 
|   |   └── image1.png
|   |   └── image2.png
|   |   └── ...                                                                                                
│   │   ├── parsings
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ... 
│   │   ├── landmarks
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ...   
│   ├── test<class1> 
|   |   └── image1.png
|   |   └── image2.png
|   |   └── ...                                                                                                
│   │   ├── parsings
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ...                                                                                                                             
...
│   ├── test<classN> 
|   |   └── image1.png
|   |   └── image2.png
|   |   └── ...                                                                                                
│   │   ├── parsings
│   │   │   └── image1.png
│   │   │   └── image2.png
│   │   │   └── ...

Citation

If you use this code for your research, please cite our paper.

Acknowledgments

This code is inspired by LATS and DLFS.

zlq1z2l3q/GGDC