#!/usr/bin/env python
from __future__ import annotations
import argparse
import pathlib
import torch
import gradio as gr
from vtoonify_model import Model
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser()
parser.add_argument('--device', type=str, default='cpu')
parser.add_argument('--theme', type=str)
parser.add_argument('--share', action='store_true')
parser.add_argument('--port', type=int)
parser.add_argument('--disable-queue',
dest='enable_queue',
action='store_false')
return parser.parse_args()
DESCRIPTION = '''
本页面是为方便英语不太好朋友了解如何使用，采用的是CPU，运算时间较长，请稳步原VToonify，速度将提升数十倍
'''
FOOTER = ''
ARTICLE = r"""
如果VToonify对你有帮助请在Github Repo上为它点亮⭐.谢谢!
[](https://github.com/williamyang1991/VToonify)
---
📝 **Citation**
If our work is useful for your research, please consider citing:
```bibtex
@article{yang2022Vtoonify,
title={VToonify: Controllable High-Resolution Portrait Video Style Transfer},
author={Yang, Shuai and Jiang, Liming and Liu, Ziwei and Loy, Chen Change},
journal={ACM Transactions on Graphics (TOG)},
volume={41},
number={6},
articleno={203},
pages={1--15},
year={2022},
publisher={ACM New York, NY, USA},
doi={10.1145/3550454.3555437},
}
```
📋 **License**
This project is licensed under S-Lab License 1.0.
Redistribution and use for non-commercial purposes should follow this license.
📧 **Contact**
If you have any questions, please feel free to reach me out at williamyang@pku.edu.cn.
"""
def update_slider(choice: str) -> dict:
if type(choice) == str and choice.endswith('-d'):
return gr.Slider.update(maximum=1, minimum=0, value=0.5)
else:
return gr.Slider.update(maximum=0.5, minimum=0.5, value=0.5)
def set_example_image(example: list) -> dict:
return gr.Image.update(value=example[0])
def set_example_video(example: list) -> dict:
return gr.Video.update(value=example[0]),
sample_video = ['./vtoonify/data/529_2.mp4','./vtoonify/data/7154235.mp4','./vtoonify/data/651.mp4','./vtoonify/data/908.mp4']
sample_vid = gr.Video(label='Video file') #for displaying the example
example_videos = gr.components.Dataset(components=[sample_vid], samples=[[path] for path in sample_video], type='values', label='Video Examples')
def main():
args = parse_args()
args.device = 'cuda' if torch.cuda.is_available() else 'cpu'
print('*** Now using %s.'%(args.device))
model = Model(device=args.device)
with gr.Blocks(theme=args.theme, css='style.css') as demo:
gr.Markdown(DESCRIPTION)
with gr.Box():
gr.Markdown('''## 第1步(选择卡通类型)
- 选择 **卡通类型**.
- 带有 `-d` 表示它可以调整卡通化的程度.
- 不带 `-d` 通常会有更好的卡通效果.
''')
with gr.Row():
with gr.Column():
gr.Markdown('''选择类型''')
with gr.Row():
style_type = gr.Radio(label='Style Type',
choices=['cartoon1','cartoon1-d','cartoon2-d','cartoon3-d',
'cartoon4','cartoon4-d','cartoon5-d','comic1-d',
'comic2-d','arcane1','arcane1-d','arcane2', 'arcane2-d',
'caricature1','caricature2','pixar','pixar-d',
'illustration1-d', 'illustration2-d', 'illustration3-d', 'illustration4-d', 'illustration5-d',
]
)
exstyle = gr.Variable()
with gr.Row():
loadmodel_button = gr.Button('加载模型')
with gr.Row():
load_info = gr.Textbox(label='Process Information', interactive=False, value='No model loaded.')
with gr.Column():
gr.Markdown('''类型参考
''')
with gr.Box():
gr.Markdown('''## 第2步 (对图片或视频进行预处理)
- 拖动1个含有人脸的图片或视频到 **输入图像**/**输入视频**.
- 点击 **缩放图像**/**缩放第1帧** 按钮.
- 缩放输入使它更好的适用模型.
- 最后的图像结果将是基于 **缩放后的脸**. 使用边框距离参数调整背景.
- **若出现[Error: no face detected!]错误**: 是因为vtoonify没有检测到人脸，请调整后再试，或者更换原始图像.
- 对于视频输入, 则点击 **缩放视频** 按钮.
- 最后的视频结果将是基于 **缩放后的视频**. 为了避免超出硬件处理能力, 视频将被裁剪成 **100/300** 帧来适应 CPU/GPU.
''')
with gr.Row():
with gr.Box():
with gr.Column():
gr.Markdown('''调整边框距离参数.
''')
with gr.Row():
top = gr.Slider(128,
256,
value=200,
step=8,
label='上')
with gr.Row():
bottom = gr.Slider(128,
256,
value=200,
step=8,
label='下')
with gr.Row():
left = gr.Slider(128,
256,
value=200,
step=8,
label='左')
with gr.Row():
right = gr.Slider(128,
256,
value=200,
step=8,
label='右')
with gr.Box():
with gr.Column():
gr.Markdown('''输入''')
with gr.Row():
input_image = gr.Image(label='输入图像',
type='filepath')
with gr.Row():
preprocess_image_button = gr.Button('缩放图像')
with gr.Row():
input_video = gr.Video(label='输入视频',
mirror_webcam=False,
type='filepath')
with gr.Row():
preprocess_video0_button = gr.Button('缩放第一帧')
preprocess_video1_button = gr.Button('绽放视频')
with gr.Box():
with gr.Column():
gr.Markdown('''View''')
with gr.Row():
input_info = gr.Textbox(label='处理信息', interactive=False, value='n.a.')
with gr.Row():
aligned_face = gr.Image(label='绽放脸',
type='numpy',
interactive=False)
instyle = gr.Variable()
with gr.Row():
aligned_video = gr.Video(label='绽放视频',
type='mp4',
interactive=False)
with gr.Row():
with gr.Column():
paths = ['./vtoonify/data/pexels-andrea-piacquadio-733872.jpg','./vtoonify/data/i5R8hbZFDdc.jpg','./vtoonify/data/yRpe13BHdKw.jpg','./vtoonify/data/ILip77SbmOE.jpg','./vtoonify/data/077436.jpg','./vtoonify/data/081680.jpg']
example_images = gr.Dataset(components=[input_image],
samples=[[path] for path in paths],
label='示例图像')
with gr.Column():
#example_videos = gr.Dataset(components=[input_video], samples=[['./vtoonify/data/529.mp4']], type='values')
#to render video example on mouse hover/click
example_videos.render()
#to load sample video into input_video upon clicking on it
def load_examples(video):
#print("****** inside load_example() ******")
#print("in_video is : ", video[0])
return video[0]
example_videos.click(load_examples, example_videos, input_video)
with gr.Box():
gr.Markdown('''## 第3步(生成 图像/视频)''')
with gr.Row():
with gr.Column():
gr.Markdown('''
- 调整 **卡通化程度**.
- 点击 **图像卡通化!** 来将第1帧卡通化. 点击 **视频卡通化!** 来让整个视频卡通化.
- 预计时间 对于300帧的1600x1440视频 : 1 小时 (CPU); 2 分钟 (GPU)
''')
style_degree = gr.Slider(0,
1,
value=0.5,
step=0.05,
label='卡通化程度')
with gr.Column():
gr.Markdown('''
''')
with gr.Row():
output_info = gr.Textbox(label='示例信息', interactive=False, value='n.a.')
with gr.Row():
with gr.Column():
with gr.Row():
result_face = gr.Image(label='图像结果',
type='numpy',
interactive=False)
with gr.Row():
toonify_button = gr.Button('图像卡通化!')
with gr.Column():
with gr.Row():
result_video = gr.Video(label='视频结果',
type='mp4',
interactive=False)
with gr.Row():
vtoonify_button = gr.Button('视频卡通化!')
gr.Markdown(ARTICLE)
gr.Markdown(FOOTER)
loadmodel_button.click(fn=model.load_model,
inputs=[style_type],
outputs=[exstyle, load_info])
style_type.change(fn=update_slider,
inputs=style_type,
outputs=style_degree)
preprocess_image_button.click(fn=model.detect_and_align_image,
inputs=[input_image, top, bottom, left, right],
outputs=[aligned_face, instyle, input_info])
preprocess_video0_button.click(fn=model.detect_and_align_video,
inputs=[input_video, top, bottom, left, right],
outputs=[aligned_face, instyle, input_info])
preprocess_video1_button.click(fn=model.detect_and_align_full_video,
inputs=[input_video, top, bottom, left, right],
outputs=[aligned_video, instyle, input_info])
toonify_button.click(fn=model.image_toonify,
inputs=[aligned_face, instyle, exstyle, style_degree, style_type],
outputs=[result_face, output_info])
vtoonify_button.click(fn=model.video_tooniy,
inputs=[aligned_video, instyle, exstyle, style_degree, style_type],
outputs=[result_video, output_info])
example_images.click(fn=set_example_image,
inputs=example_images,
outputs=example_images.components)
demo.launch(
enable_queue=args.enable_queue,
server_port=args.port,
share=args.share,
)
if __name__ == '__main__':
main()