Home

Awesome

$E^3$-FaceNet

This is an official implementation of ICML 2024 Paper "Fast Text-to-3D-Aware Face Generation and Manipulation via Direct Cross-modal Mapping and Geometric Regularization". The proposed $E^3$-FaceNet is an End-to-End Efficient and Effective network for fast and accurate T3D face generation and manipulation, which can not only achieve picture-like 3D face generation and manipulation, but also improve inference speed by orders of magnitudes.

🚀 Overview

🖥️ Setup

Environment

The codebase is tested on

For additional python libraries, please install by:

pip install -r requirements.txt

Please refer to https://github.com/NVlabs/stylegan2-ada-pytorch for additional software/hardware requirements.

[!TIP] A modification has been made to the clip package to enable simultaneous extraction of text features and token embeddings. Please replace the existing model.py file in your own clip installation path.

Data Preparation

We train our $E^3$-FaceNet on Multi-Modal-CelebA-HQ Dataset (MM-CelebA) and evaluation on MMCelebA, CelebAText-HQ and FFHQ-Text.

Before training, please dowload the dataset2.json, and place the file in the MMceleba dataset directory.

Pretrained Checkpoints

The model weight can be download at here.

Training $E^3$-FaceNet

use the shell script,

bash train_train_4_E3_Face.sh

Please check configuration files at conf/model and conf/spec. You can always add your own model config. More details on how to use hydra configuration please follow https://hydra.cc/docs/intro/.

Evaluate $E^3$-FaceNet

use the shell script,

bash run_eval_4_E3_Face.sh

Text-Guided Generation and Manipulation

use the shell script,

bash sample.sh

Visual Results

Compare with Text-to-3D Face Methods

<img src="figure/3D-readme.png" style="zoom:67%;" />

Compare with Text-to-2D Face Methods

<img src="figure/2D-readme.png" style="zoom: 67%;" />

🖊️ Citation

If $E^3$-FaceNet is helpful for your research or you wish to refer the baseline results published here, we'd really appreciate it if you could cite this paper:

@InProceedings{zhang2024fast,
  title = 	 {Fast Text-to-3{D}-Aware Face Generation and Manipulation via Direct Cross-modal Mapping and Geometric Regularization},
  author =       {Zhang, Jinlu and Zhou, Yiyi and Zheng, Qiancheng and Du, Xiaoxiong and Luo, Gen and Peng, Jun and Sun, Xiaoshuai and Ji, Rongrong},
  booktitle = 	 {Proceedings of the 41st International Conference on Machine Learning},
  pages = 	 {60605--60625},
  year = 	 {2024},
  editor = 	 {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix},
  volume = 	 {235},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {21--27 Jul},
  publisher =    {PMLR},
  pdf = 	 {https://raw.githubusercontent.com/mlresearch/v235/main/assets/zhang24cp/zhang24cp.pdf},
  url = 	 {https://proceedings.mlr.press/v235/zhang24cp.html},
}

🎫 Acknowledgment

This project largely references StyleNeRF. Thanks for their amazing work!