Awesome

Practical-RIFE

V4.0 Anime Demo Video | 迭代经验 | Colab

This project is based on RIFE and SAFA. We aim to enhance their practicality for users by incorporating various features and designing new models. Since improving the PSNR index is not consistent with subjective perception. This project is intended for engineers and developers. For general users, we recommend the following software:

SVFI (中文) | RIFE-App | FlowFrames

Thanks to SVFI team to support model testing on Animation.

VapourSynth-RIFE | RIFE-ncnn-vulkan | VapourSynth-RIFE-ncnn-Vulkan | vs-mlrt | Drop frame fixer and FPS converter

Frame Interpolation

2024.08 - We find that 4.24+ is quite suitable for post-processing of some diffusion model generated videos.

Trained Model

The content of these links is under the same MIT license as this project. lite means using similar training framework, but lower computational cost model. Currently, it is recommended to choose 4.25 by default for most scenes.

4.26 - 2024.09.21 | Google Drive 百度网盘 || 4.26.heavy || 4.25.lite - 2024.10.20

4.25 - 2024.09.19 | Google Drive 百度网盘 | I am trying using more flow blocks, so the scale_list will change accordingly. It seems that the anime scenes have been significantly improved.

4.22 - 2024.08.08 | Google Drive 百度网盘 || 4.22.lite || 4.21 - 2024.08.04 | Google Drive 百度网盘

4.20 - 2024.07.24 | Google Drive 百度网盘 || 4.18 - 2024.07.03 | Google Drive 百度网盘

4.17 - 2024.05.24 | Google Drive 百度网盘 : Add gram loss from FILM || 4.17.lite

4.15 - 2024.03.11 | Google Drive 百度网盘 || 4.15.lite || 4.14 - 2024.01.08 | Google Drive 百度网盘 || 4.14.lite

v4.9.2 - 2023.11.01 | Google Drive 百度网盘 || v4.3 - 2022.8.17 | Google Drive 百度网盘

v3.8 - 2021.6.17 | Google Drive 百度网盘 || v3.1 - 2021.5.17 | Google Drive 百度网盘

More Older Version

Installation

python <= 3.11

git clone git@github.com:hzwer/Practical-RIFE.git
cd Practical-RIFE
pip3 install -r requirements.txt

Download a model from the model list and put *.py and flownet.pkl on train_log/

Run

You can use our demo video or your video.

python3 inference_video.py --multi=2 --video=video.mp4 

(generate video_2X_xxfps.mp4)

python3 inference_video.py --multi=4 --video=video.mp4

(for 4X interpolation)

python3 inference_video.py --multi=2 --video=video.mp4 --scale=0.5

(If your video has high resolution, such as 4K, we recommend set --scale=0.5 (default 1.0))

python3 inference_video.py --multi=4 --img=input/

(to read video from pngs, like input/0.png ... input/612.png, ensure that the png names are numbers)

Parameter descriptions:

--img / --video: The input file address

--output: Output video name 'xxx.mp4'

--model: Directory with trained model files

--UHD: It is equivalent to setting scale=0.5

--montage: Splice the generated video with the original video, like this demo

--fps: Set output FPS manually

--ext: Set output video format, default: mp4

--multi: Interpolation frame rate multiplier

--exp: Set --multi to 2^(--exp)

--skip: It's no longer useful refer to issue 207

Model training

The whole repo can be downloaded from v4.0, v4.12, v4.15, v4.18. However, we currently do not have the time to organize them well, they are for reference only.

Video Enhancement

We are developing a practical model of SAFA. Welcome to check its demo (BiliBili) and provide advice.

v0.5 - 2023.12.26 | Google Drive

python3 inference_video_enhance.py --video=demo.mp4

Citation

@inproceedings{huang2022rife,
  title={Real-Time Intermediate Flow Estimation for Video Frame Interpolation},
  author={Huang, Zhewei and Zhang, Tianyuan and Heng, Wen and Shi, Boxin and Zhou, Shuchang},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  year={2022}
}

@inproceedings{huang2024safa,
  title={Scale-Adaptive Feature Aggregation for Efficient Space-Time Video Super-Resolution},
  author={Huang, Zhewei and Huang, Ailin and Hu, Xiaotao and Hu, Chen and Xu, Jun and Zhou, Shuchang},
  booktitle={Winter Conference on Applications of Computer Vision (WACV)},
  year={2024}
}

Reference

Optical Flow: ARFlow pytorch-liteflownet RAFT pytorch-PWCNet

Video Interpolation: DVF TOflow SepConv DAIN CAIN MEMC-Net SoftSplat BMBC EDSC EQVI RIFE