Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis

This is the official repository for our ICCV 2023 paper Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis.

Paper | Project | ArXiv | Video

Update

• [2024/07/02] Our new work TalkingGaussian is released!
• TODO: Use AU to implement SyncTalk's full expression control, just like what we do in TalkingGaussian.

Installation

Tested on Ubuntu 18.04, Pytorch 1.12 and CUDA 11.3.

Install dependency

conda create -n ernerf python=3.10
conda install pytorch==1.12.1 torchvision==0.13.1 cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt
pip install "git+https://github.com/facebookresearch/pytorch3d.git"
pip install tensorflow-gpu==2.8.0

Preparation

• Prepare face-parsing model.

  wget https://github.com/YudongGuo/AD-NeRF/blob/master/data_util/face_parsing/79999_iter.pth?raw=true -O data_utils/face_parsing/79999_iter.pth

• Prepare the 3DMM model for head pose estimation.

  wget https://github.com/YudongGuo/AD-NeRF/blob/master/data_util/face_tracking/3DMM/exp_info.npy?raw=true -O data_utils/face_tracking/3DMM/exp_info.npy
  wget https://github.com/YudongGuo/AD-NeRF/blob/master/data_util/face_tracking/3DMM/keys_info.npy?raw=true -O data_utils/face_tracking/3DMM/keys_info.npy
  wget https://github.com/YudongGuo/AD-NeRF/blob/master/data_util/face_tracking/3DMM/sub_mesh.obj?raw=true -O data_utils/face_tracking/3DMM/sub_mesh.obj
  wget https://github.com/YudongGuo/AD-NeRF/blob/master/data_util/face_tracking/3DMM/topology_info.npy?raw=true -O data_utils/face_tracking/3DMM/topology_info.npy

• Download 3DMM model from Basel Face Model 2009:

  # 1. copy 01_MorphableModel.mat to data_util/face_tracking/3DMM/
  # 2.
    cd data_utils/face_tracking
    python convert_BFM.py
  

Datasets and pretrained models

We get the experiment videos mainly from AD-NeRF, DFRF, GeneFace and YouTube. Due to copyright restrictions, we can't distribute all of them. You may have to download and crop these videos by youself. Here is an example training video (Obama) from AD-NeRF with the resolution of 450x450.

mkdir -p data/obama
wget https://github.com/YudongGuo/AD-NeRF/blob/master/dataset/vids/Obama.mp4?raw=true -O data/obama/obama.mp4

We also provide pretrained checkpoints on the Obama video clip. After completing the data pre-processing step, you can download and test them by:

python main.py data/obama/ --workspace trial_obama/ -O --test --ckpt trial_obama/checkpoints/ngp.pth   # head
python main.py data/obama/ --workspace trial_obama_torso/ -O --test --torso --ckpt trial_obama_torso/checkpoints/ngp.pth   # head+torso

The test results should be about:

setting	PSNR	LPIPS	LMD
head	35.607	0.0178	2.525
head+torso	26.594	0.0446	2.550

Usage

Pre-processing Custom Training Video

• Put training video under data/<ID>/<ID>.mp4.

  The video must be 25FPS, with all frames containing the talking person. The resolution should be about 512x512, and duration about 1-5 min.

• Run script to process the video. (may take several hours)

  python data_utils/process.py data/<ID>/<ID>.mp4

• Obtain AU45 for eyes blinking

  Run FeatureExtraction in OpenFace, rename and move the output CSV file to data/<ID>/au.csv.

Audio Pre-process

In our paper, we use DeepSpeech features for evaluation.

You should specify the type of audio feature by --asr_model <deepspeech, esperanto, hubert> when training and testing.

• DeepSpeech

  python data_utils/deepspeech_features/extract_ds_features.py --input data/<name>.wav # save to data/<name>.npy

• Wav2Vec

  You can also try to extract audio features via Wav2Vec like RAD-NeRF by:

  python data_utils/wav2vec.py --wav data/<name>.wav --save_feats # save to data/<name>_eo.npy

• HuBERT

  In our test, HuBERT extractor performs better for more languages, which has already been used in GeneFace.

  # Borrowed from GeneFace. English pre-trained.
  python data_utils/hubert.py --wav data/<name>.wav # save to data/<name>_hu.npy

Train

First time running will take some time to compile the CUDA extensions.

# train (head and lpips finetune, run in sequence)
python main.py data/obama/ --workspace trial_obama/ -O --iters 100000
python main.py data/obama/ --workspace trial_obama/ -O --iters 125000 --finetune_lips --patch_size 32

# train (torso)
# <head>.pth should be the latest checkpoint in trial_obama
python main.py data/obama/ --workspace trial_obama_torso/ -O --torso --head_ckpt <head>.pth --iters 200000

Test

# test on the test split
python main.py data/obama/ --workspace trial_obama/ -O --test # only render the head and use GT image for torso
python main.py data/obama/ --workspace trial_obama_torso/ -O --torso --test # render both head and torso

Inference with target audio

# Adding "--smooth_path" may help decrease the jitter of the head, while being less accurate to the original pose.
python main.py data/obama/ --workspace trial_obama_torso/ -O --torso --test --test_train --aud <audio>.npy

Citation

Consider citing as below if you find this repository helpful to your project:

@InProceedings{li2023ernerf,
    author    = {Li, Jiahe and Zhang, Jiawei and Bai, Xiao and Zhou, Jun and Gu, Lin},
    title     = {Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {7568-7578}
}

Acknowledgement

This code is developed heavily relying on RAD-NeRF, and also DFRF, GeneFace, and AD-NeRF. Thanks for these great projects.