first commit

.gitignore

@@ -0,0 +1,4 @@
+video_split/
+video_split_swap/
+test.ipynb
+test.py

README.md

@@ -0,0 +1,59 @@
+# 🎭 Painted Skin 🎭
+
+> Welcome to the 🎭 Painted Skin 🎭! Get ready to have some fun with faces. 😄
+
+ ![License](https://img.shields.io/badge/license-Apache 2.0-blue)
+
+## 🐾 Preview
+
+![image-20231108160501352](https://black-thompson.oss-cn-beijing.aliyuncs.com/img/image-20231108160501352.png)
+
+
+
+## 🪄 How to use
+
+1. **Clone the Repository:**
+
+    ```bash
+    git clone https://github.com/BlackThompson/Painted-Skin.git
+    ```
+
+2. **Create a New Environment and Install Dependencies:**
+
+    - Note: Python version should be >=3.8
+    - Note: The versions of `torch`, `torchvision`, and `torchaudio` should align with your CUDA version.
+
+    ```bash
+    conda create --name painted_skin python=3.8
+    conda activate painted_skin
+    pip install -r requirements.txt 
+    ```
+
+3. **Run `gradio_UI.py`:**
+    ```bash
+    gradio gradio_UI.py
+    ```
+
+4. **Click the Local URL**
+
+## 🦾 How it Works
+
+This tool allows you to perform a face swap. Simply follow these steps:
+
+1. **Upload Source:**
+   - Choose between `src_picture` and `src_video` for the face you want to swap.
+   - Note: You can only upload one source at a time, and the tool processes one task at a time.
+2. **Upload Target:**
+   - Upload `target_picture`, and our model will swap the face onto the source image or video.
+3. **Submit and Wait:**
+   - Click the submit button and patiently wait for the process to complete.
+   - Keep in mind that face swapping in videos might take some time, so be patient! 😅
+
+## ❗ Important Note
+- Ensure that `target_picture` is uploaded for the face swap to work effectively.
+
+Enjoy swapping faces and have a good laugh! 😆
+
+## 💌 Acknowledgements
+
+This repository borrows heavily from [facefusion-damo](https://www.modelscope.cn/models/damo/cv_unet-image-face-fusion_damo/summary) and [face-change](https://github.com/Quietbe/mv_face_change/blob/main/video_cut_cv_h.py). Thanks to the authors for sharing their code and models.

face_swap_utils.py

@@ -0,0 +1,161 @@
+# _*_ coding : utf-8 _*_
+# @Time : 2023/11/6 17:04
+# @Author : Black
+# @File : face_swap_utils
+# @Project : face_fusion_damo
+
+import cv2
+import os
+from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_audio
+import moviepy.editor as mp
+import torch
+from modelscope.outputs import OutputKeys
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+from moviepy.editor import VideoFileClip, AudioFileClip, ImageSequenceClip
+import natsort
+
+
+# 将视频video_path分割成图片和音频文件，保存到save_path文件夹中
+def video2mp3_img(video_path, save_path):
+    def _video2img(video_path, save_path):
+        cap = cv2.VideoCapture(video_path)
+        frame_rate = int(cap.get(cv2.CAP_PROP_FPS))  # 获取视频帧率
+        i = 0
+        while True:
+            ret, frame = cap.read()
+            if ret:
+                cv2.imwrite(save_path + '/' + str(i) + '.jpg', frame)
+                i += 1
+            else:
+                break
+        cap.release()
+        print(f"Video frame rate: {frame_rate}")
+        return frame_rate
+
+    def _video2mp3(video_path, save_path):
+        # 提取音频并保存为临时文件（默认是WAV格式）
+        audio_temp_file = os.path.join(save_path, 'audio.wav')
+        ffmpeg_extract_audio(video_path, audio_temp_file)
+
+        # 将音频转换为MP3格式
+        audio_clip = mp.AudioFileClip(audio_temp_file)
+        audio_temp_file = os.path.join(save_path, 'audio.mp3')
+        audio_clip.write_audiofile(audio_temp_file)
+        # 删除临时音频文件
+        audio_clip.close()
+
+    if not os.path.exists(save_path):
+        os.makedirs(save_path)
+
+    # 视频分割
+    frame_rate = _video2img(video_path, save_path)
+    print("split picture finished!")
+    # 视频提取音频
+    _video2mp3(video_path, save_path)
+    print("extract audio finished!")
+
+    return frame_rate
+
+
+# 将图片文件夹中的图片进行人脸替换，保存到save_dir文件夹中
+def replace_all_img(template_dir, user_path, save_dir):
+    # image_face_fusion = pipeline(Tasks.image_face_fusion, model='damo/cv_unet-image-face-fusion_damo')
+    for root, dirs, files in os.walk(template_dir):
+        for file in files:
+            if file.endswith(".jpg") or file.endswith(".png"):
+                template_path = os.path.join(root, file)
+                replace_single_img(template_path=template_path, user_path=user_path,
+                                   save_dir=save_dir)
+
+
+# 将单张图片进行人脸替换，保存到save_dir文件夹中
+def replace_single_img(template_path, user_path, save_dir):
+    image_face_fusion = pipeline(Tasks.image_face_fusion, model='damo/cv_unet-image-face-fusion_damo')
+    filename = os.path.basename(template_path)
+    print(f"{filename} start face swapping!")
+    # filename = os.path.splitext(os.path.basename(template_path))[0]
+
+    # 替换面部依赖: template为原图,即视频拆分图; user为用户要替换脸的图
+    result = image_face_fusion(dict(template=template_path, user=user_path))
+    filepath = os.path.join(save_dir, filename)
+    cv2.imwrite(filepath, result[OutputKeys.OUTPUT_IMG])
+
+    # 释放CUDA内存
+    # 只是删除了该内部函数作用域中的变量引用，而不会影响外部函数的变量
+    del image_face_fusion
+    torch.cuda.empty_cache()
+    print(f"{filename} finished!")
+    return filepath
+
+
+# 将图片文件夹中的图片合成视频，保存到output_dir文件夹中
+def img2mp4(save_name, output_dir, img_folder, mp3_folder, fps=25):
+    images = []
+    for root, dirs, files in os.walk(img_folder):
+        for file in files:
+            if file.endswith(".jpg") or file.endswith(".png"):
+                file_path = os.path.join(root, file)
+                images.append(file_path)
+    # 将图片按照文件名进行自然排序
+    images = natsort.natsorted(images)
+
+    clips = [ImageSequenceClip(images, fps=fps)]
+    video_clip = clips[0]
+    audio_path = os.path.join(mp3_folder, 'audio.wav')
+    audio_clip = AudioFileClip(audio_path)
+
+    # 如果音频和视频时长不匹配，可以选择截取或重复音频以使其匹配视频长度
+    if audio_clip.duration > video_clip.duration:
+        audio_clip = audio_clip.subclip(0, video_clip.duration)
+    else:
+        video_clip = video_clip.subclip(0, audio_clip.duration)
+
+    video_clip = video_clip.set_audio(audio_clip)
+
+    save_name = save_name + '.mp4'
+    video_clip.write_videofile(os.path.join(output_dir, save_name), codec="libx264")
+    return os.path.join(output_dir, save_name)
+
+
+# 整个视频的替换人脸
+def replace_video(video_path, user_path):
+    # .py文件所在路径
+    BASE = os.path.dirname(__file__)
+    template_dir = os.path.join(BASE, 'video_split')
+    save_dir = os.path.join(BASE, 'video_split_swap')
+    output_dir = os.path.join(BASE, 'output')
+
+    if not os.path.exists(template_dir):
+        os.makedirs(template_dir)
+    if not os.path.exists(save_dir):
+        os.makedirs(save_dir)
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+
+    # 将视频分割成图片和音频
+    fps = video2mp3_img(video_path=video_path, save_path=template_dir)
+    replace_all_img(template_dir=template_dir, user_path=user_path, save_dir=save_dir)
+    save_path = img2mp4(save_name='swapped_video', output_dir=output_dir, img_folder=save_dir, mp3_folder=template_dir,
+                        fps=fps)
+    return save_path
+
+
+if __name__ == '__main__':
+    # .py文件所在路径
+    BASE = os.path.dirname(__file__)
+    # .ipynb文件所在路径
+    # BASE = os.getcwd()
+    # 设置工作路径为当前脚本所在的路径
+    os.chdir(BASE)
+
+    user_path = r'./input/Trump.jpg'
+    template_dir = r'./middle'
+    save_dir = r'./middle_after_swap'
+    output_dir = r'./output'
+    video_path = r'./input/zhan.mp4'
+
+    # 将视频分割成图片和音频
+    fps = video2mp3_img(video_path=video_path, save_path=template_dir)
+    replace_all_img(template_dir=template_dir, user_path=user_path, save_dir=save_dir)
+    img2mp4(save_name='Zhan_Trump', output_dir=output_dir, img_folder=save_dir, mp3_folder=template_dir, fps=fps)

gradio_UI.py

@@ -0,0 +1,64 @@
+# _*_ coding : utf-8 _*_
+# @Time : 2023/11/7 21:42
+# @Author : Black
+# @File : gradio_UI
+# @Project : face_fusion_damo
+
+import gradio as gr
+import os
+from face_swap_utils import *
+
+
+def face_swap(src_picture=None, src_video=None,
+              target_picture=os.path.join(os.path.dirname(__file__), 'example/Trump.jpg')):
+    save_dir = os.path.join(os.path.dirname(__file__), 'output')
+    # 检查是否存在临时文件夹，如果不存在则创建
+    if not os.path.exists(save_dir):
+        os.makedirs(save_dir)
+
+    if src_picture is not None:
+        print(src_picture)
+        result = replace_single_img(src_picture, target_picture, save_dir)
+        return result, None
+    elif src_video is not None:
+        print(src_video)
+        result = replace_video(src_video, target_picture)
+        return None, result
+
+
+demo = gr.Interface(fn=face_swap,
+                    inputs=[gr.Image(type='filepath'), gr.Video(), gr.Image(type='filepath')],
+                    outputs=[gr.Image(type='filepath', label='swapped_image'),
+                             gr.Video(label='swapped_video')],
+                    title="🎭 Painted Skin 🎭",
+                    description="# Face Swap Tool\n"
+                                "Welcome to the 🎭 Painted Skin 🎭! Get ready to have some fun with faces. 😄\n"
+                                "## How it Works\n"
+                                "This tool allows you to perform a face swap. Simply follow these steps:\n"
+                                "1. **Upload Source:**\n"
+                                "   - Choose between `src_picture` and `src_video` for the face you want to swap.\n"
+                                "   - Note: You can only upload one source at a time, and the tool processes one task at a time.\n"
+                                "2. **Upload Target:**\n"
+                                "   - Upload `target_picture`, and our model will swap the face onto the source image or video.\n"
+                                "3. **Submit and Wait:**\n"
+                                "   - Click the submit button and patiently wait for the process to complete.\n"
+                                "   - Keep in mind that face swapping in videos might take some time, so be patient! 😅\n"
+                                "## Important Note\n"
+                                "- Ensure that `target_picture` is uploaded for the face swap to work effectively.\n"
+                                "Enjoy swapping faces and have a good laugh! 😆\n",
+
+                    # css="body {background-color: red;}",
+                    # article="Attention is all you need!",
+                    examples=[[os.path.join(os.path.dirname(__file__), 'example/Trump.jpg'),
+                               None,
+                               os.path.join(os.path.dirname(__file__), 'example/Obama.jpg')],
+                              [os.path.join(os.path.dirname(__file__), 'example/CXK.jpg'),
+                               None,
+                               os.path.join(os.path.dirname(__file__), 'example/James.jpg')]
+                              ],
+                    # allow_flagging="never",
+                    # cache_examples=True
+                    )
+
+if __name__ == "__main__":
+    demo.launch(share=True)

requirements.txt

@@ -0,0 +1,6 @@
+gradio==4.1.2
+modelscope==1.9.4
+moviepy==1.0.3
+natsort==8.4.0
+opencv_python==4.8.1.78
+torch==1.13.1+cu116