Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[IJCAI'24] Beyond Alignment: Blind Video Face Restoration via Parsing-Guided Temporal-Coherent Transformer
| Date | Stars |
|---|---|
| 2026-07-24 | 689 |
| 2026-07-25 | 689 |
| 2026-07-28 | 689 |
| 2026-07-30 | 689 |
| 2026-08-06 | 689 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
<div align="center">
<h1>Beyond Alignment: Blind Video Face Restoration via Parsing-Guided Temporal-Coherent Transformer</h1>
<div>
<a href='https://kepengxu.github.io/' target='_blank'>Kepeng Xu</a> 
<a href='https://icecherylxuli.github.io/' target='_blank'>Li Xu</a> 
<a href='' target='_blank'>Gang He et al</a>
</div>
<div>
Xidian University,Southwest University of Science and Technology 
</div>
<div>
<strong>IJCAI 2024</strong>
</div>
<div>
<h4 align="center">
<a href="https://kepengxu.github.io/projects/pgtformer/" target='_blank'>
<img src="https://img.shields.io/badge/🐳-Project%20Page-blue">
</a>
</h4>
</div>
<p align="center">
<img src="./assets/output.gif" alt="showcase">
<br>
🔥 For more results, visit our <a href="https://kepengxu.github.io/projects/pgtformer/"><strong>project page</strong>,<a href="https://github.com/kepengxu/PGTFormer/blob/main/MoreVideoResults.md"><strong>gif page</strong></a> 🔥
<br>
⭐ If you found this project helpful to your projects, please help star this repo. Thanks! 🤗
</p>
</div>
if you have any questions; please send message to me via Wechat or email: [email protected]
<p align="center">
<img src="./assets/vx.png" alt="Wechat" width="200">
</p>
# Update
- **2024.09**: Update Windows inference script.(Thanks @KFK121)
- **2024.08**: 🤗 This model has been successfully pushed to the Hugging Face **kepeng/pgtformer-base**, making it accessible to researchers and developers worldwide. Feel free to explore, test, and utilize it
- **2024.08**: We released the initial version of the inference code and models. Stay tuned for continuous updates!
- **2024.04**: This repo is created!
# Getting Started
## Dependencies and Installation
required packages in `requirements`
```
# git clone this repository
git clone https://github.com/kepengxu/PGTFormer
cd PGTFormer
# create new anaconda env
conda create -n pgtformer python=3.8 -y
conda activate pgtformer
# install python dependencies
conda install -c conda-forge dlib
conda install -c conda-forge ffmpeg
```
## Quick Inference
### Download Pre-trained Models
All pretrained models can also be automatically downloaded during the first inference.
You can also download our pretrained models from [Google Drive](https://drive.google.com/file/d/1DFwfPpiIxqjd-PrQ7zoKlVF4HqTM0Ops/view?usp=sharing),[BaiduYun](https://pan.baidu.com/s/1pN0u6ITT-JUFg9-PhgUkrg?pwd=pgtf),[Huggingface](https://huggingface.co/kepeng/pgtformer-base) to the `weights` folder.
### Prepare Testing Data
We provide a example in `assets/inputdemovideo.mp4`. If you would like to test your own face videos, place them in the same folder.
### Inference
**[Note]** 🚀🚀🚀 Our method does **not require pre-alignment** to standard face poses and has better consistency.
The results will be saved in the `results` folder.
🧑🏻 Video Face Restoration
**It is highly recommended to use ffmpeg to enhance the encoded video online, which can maintain the restored quality as much as possible.**
```
# Just Run This
# Input Video Width and Height == 512
mkdir exp
python inference.py --input_video=assets/inputdemovideo.mp4 --output_video=exp/output_demo.mp4
```
### 🚀🚀🚀Awesome Video Face Restoration Method Can be Found There!🚀🚀🚀
[Awesome Video Face Restoration](https://github.com/kepengxu/Awesome-Video-Face-Restoration/tree/main)
# 🔥 News
- *2024.04*: 🎉 This paper is accepted by IJCAI 2024
- Beyond Alignment: Blind Video Face Restoration via Parsing-Guided Temporal-Coherent Transformer (IJCAI 2024)
### Results report in VFHQ implementation.
| Metrics | Bicubic | EDVRM | BasicVSR | EDVRM-GAN | BasicVSR-GAN | DFDNet | GFP-GAN | GPEN |
|---------|---------|-------|----------|-----------|--------------|--------|---------|------|
| PSNR | 26.842 | 29.457 (Red) | 29.472 (Red) | 26.682 | 25.813 | 25.178 | 25.978 | 26.672 |
| SSIM | 0.7909 | 0.8428 (Red) | 0.8430 (Red) | 0.7638 | 0.741 | 0.7560 | 0.7723 Excerpt of 6,042 characters
Read on GitHub
26
5
HuggingFace · Belgium
2
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:7906ea5e82c05b3c, topic:super-resolution