Adaptive Super Resolution For One-Shot Talking-Head Generation

The repository for ICASSP2024 Adaptive Super Resolution For One-Shot Talking-Head Generation (AdaSR TalkingHead)

Abstract

The one-shot talking-head generation learns to synthesize a talking-head video with one source portrait image under the driving of same or different identity video. Usually these methods require plane-based pixel transformations via Jacobin matrices or facial image warps for novel poses generation. The constraints of using a single image source and pixel displacements often compromise the clarity of the synthesized images. Some methods try to improve the quality of synthesized videos by introducing additional super-resolution modules, but this will undoubtedly increase computational consumption and destroy the original data distribution. In this work, we propose an adaptive high-quality talking-head video generation method, which synthesizes high-resolution video without additional pre-trained modules. Specifically, inspired by existing super-resolution methods, we down-sample the one-shot source image, and then adaptively reconstruct high-frequency details via an encoder-decoder module, resulting in enhanced video clarity. Our method consistently improves the quality of generated videos through a straightforward yet effective strategy, substantiated by quantitative and qualitative evaluations. The code and demo video are available on: https://github.com/Songluchuan/AdaSR-TalkingHead/

Updates

[03/2024] Inference code and pretrained model are released.
[03/2024] Arxiv Link: https://arxiv.org/abs/2403.15944.
[COMING] Super-resolution model (based on StyleGANEX and ESRGAN).
[COMING] Train code and processed datasets.

Installation

Clone this repo:

git clone git@github.com:Songluchuan/AdaSR-TalkingHead.git
cd AdaSR-TalkingHead

Dependencies:

We have tested on:

CUDA 11.3-11.6
PyTorch 1.10.1
Matplotlib 3.4.3; Matplotlib 3.4.2; opencv-python 4.7.0; scikit-learn 1.0; tqdm 4.62.3

Inference Code

Download the pretrained model on google drive: https://drive.google.com/file/d/1g58uuAyZFdny9_twvbv0AHxB9-03koko/view?usp=sharing (it is trained on the HDTF dataset), and put it under checkpoints/
The demo video and reference image are under DEMO/
The inference code is in the run_demo.sh, please run it with

bash run_demo.sh

You can set different demo image and driven video in the run_demo.sh

--source_image DEMO/demo_img_3.jpg

and

--driving_video DEMO/demo_video_1.mp4

Video

Citation

@inproceedings{song2024adaptive,
  title={Adaptive Super Resolution for One-Shot Talking Head Generation},
  author={Song, Luchuan and Liu, Pinxin and Yin, Guojun and Xu, Chenliang},
  year={2024},
  organization={IEEE International Conference on Acoustics, Speech, and Signal Processing}
}

Acknowledgments

The code is mainly developed based on styleGANEX, ESRGAN and unofficial face2vid. Thanks to the authors contribution.