Skip to content

Latest commit

 

History

History
282 lines (208 loc) · 15.3 KB

README.md

File metadata and controls

282 lines (208 loc) · 15.3 KB
Raw

AnimeGANv3

Paper Title: A Novel Double-Tail Generative Adversarial Network for Fast Photo Animation.

Let's use AnimeGANv3 to produce our own animation.

manuscript Paper Project Page HuggingFace Video twitter LICENSE Github Open In Colab Visitor

📢 Updates

  • 2024-12-15 Released the rknn-based model conversion and inference repo.
  • 2024-11-27 Disney and Trump styles updated to version 2.0.
  • 2024-10-31 Added a new styles of AnimeGANv3: Portrait to Pixar. 🎃
  • 2024-08-28 A repo more suitable for portrait style inference based on the AnimeGANv3 models has been released. Highly recommended.🔥
  • 2023-12-10 Added a new AnimeGANv3 model for Portrait to Oil-painting style. Its onnx is available here.
  • 2023-11-23 The code and the manuscript are released. 🦃
  • 2023-10-31 Added three new styles of AnimeGANv3: Portrait to Cute, 8bit and Sketch-0 style. 👻
  • 2023-09-18 Added a new AnimeGANv3 model for Face to Kpop style.
  • 2023-01-16 Added a new AnimeGANv3-photo.exe for the inference of AnimeGANv3's onnx model.
  • 2023-01-13 Added a new AnimeGANv3 model for Face to comic style.
  • 2022-12-25 Added the tiny model (2.4MB) of Nordic myth style and USA style 2.0. It can go upto 50 FPS on iphone14 with 512*512 input. 🎅
  • 2022-11-24 Added a new AnimeGANv3 model for Face to Nordic myth style. 🦃
  • 2022-11-06 Added a new AnimeGANv3 model for Face to Disney style V1.0.
  • 2022-10-31 Added a new AnimeGANv3 model for Face to USA cartoon and Disney style V1.0. 🎃
  • 2022-10-07 The USA cartoon Style of AnimeGANv3 is integrated to ProfileProfile with Core ML. Install it by the Apple Store and have a try.
  • 2022-09-26 Official online demo is integrated to Huggingface Spaces with Gradio. Hugging Face Spaces
  • 2022-09-24 Added a new great AnimeGANv3 model for Face to USA cartoon Style.
  • 2022-09-18 Update a new AnimeGANv3 model for Photo to Hayao Style.
  • 2022-08-01 Added a new AnimeGANv3 onnx model (Colab) for Face to Arcane style.
  • 2022-07-13 Added a new AnimeGANv3 onnx model (Colab) for Face to portrait sketch.
  • 2021-12-25 The paper of AnimeGANv3 will be released in 2022. 🎄

🎮 Usage

  • Official online demo is released in Hugging Face Spaces.

  • Download this repository and use AnimeGANv3's UI tool and pre-trained *.onnx to turn your photos into anime. 😊

  • 🛠️ Installation

    1. Clone repo

      git clone https://github.com/TachibanaYoshino/AnimeGANv3.git
cd AnimeGANv3

    2. Install dependent packages

      pip install -r requirements.txt

    3. Inference with *.onnx

      python deploy/test_by_onnx.py -i inputs/imgs/ -o output/results -m deploy/AnimeGANv3_Hayao_36.onnx

    4. video to anime with *.onnx

      python tools/video2anime.py -i inputs/vid/1.mp4 -o output/results -m deploy/AnimeGANv3_Hayao_36.onnx

🚀 Landscape Demos

🔥 Video to anime (Hayao Style)

🎨 Photo to Hayao Style


more surprise 👈






🎨 Photo to Shinkai Style


more surprise 👈





🚀 Portrait Style Demos

The paper has been completed in 2022. The study of portrait stylization is an extension of the paper.

Some exhibits  👈

🎨 Face to USA cartoon style

8_USA.mp4

🎨 Face to Disney cartoon style

v1.9 v2.0
x_sound.Disney-v1.9.mp4
x_sound.Disney2.0.mp4

🎨 Face to Trump style

v1.9 v2.0
20_Trump-v1.9.mp4
20_Trump-v2.0.mp4
more surprise 👈

🎨 Face to Arcane style

10c.mp4

🎨 Portrait to comic style

16_sound.mp4

🎨 Face to Kpop style

6c-Kpop.mp4

🎨 Portrait to Oil-painting style

more surprise 👈

🎨 Portrait to Cute style

5_sound.Cute.mp4

🎨 Portrait to Pixar style

12_AnimeGANv3_Pixar_sounds.mp4

🎨 Portrait to Sketch-0 style

11_AnimeGANv3_light_Sketch-0_soundsc6.mp4

🎨 Portrait to 8bit style

🎨 Face to portrait sketch

Open In Colab

input Face panoramic image
more surprise 👈


🔨 Train

1. Download dataset and pretrained vgg19

  1. vgg19
  2. Hayao dataset
  3. Shinkai dataset
  4. photo dataset

2. Do edge_smooth

    cd tools && python edge_smooth.py --dataset Hayao --img_size 256

3. Do superPixel

    cd tools && python visual_superPixel_seg_image.py

4. Train

    python train.py --style_dataset Hayao --init_G_epoch 5 --epoch 100

✒️ Citation

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

@article{Liu2024dtgan,
  title={A Novel Double-Tail Generative Adversarial Network for Fast Photo Animation},
  author={Gang LIU and Xin CHEN and Zhixiang GAO},
  journal={IEICE Transactions on Information and Systems},
  volume={E107.D},
  number={1},
  pages={72-82},
  year={2024},
  doi={10.1587/transinf.2023EDP7061}
}

📜 License

This repo is made freely available to academic and non-academic entities for non-commercial purposes such as academic research, teaching, scientific publications. Permission is granted to use the AnimeGANv3 given that you agree to my license terms. Regarding the request for commercial use, please contact us via email to help you obtain the authorization letter.

📧 Author

Asher Chan [email protected]