Hugging Face's logo

Spaces:
FolSpark
/
FakeVideoDetect
Build error

App Files Community
FakeVideoDetect
File size: 2,213 Bytes 
b0c0f32
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
 
# # 使用moviepy从视频中分离音频

# from moviepy.editor import VideoFileClip

# # 定义输入视频文件的路径
# video_file_path = '/mnt/data10t/dazuoye/GROUP2024-GEN6/FakeSV/dataset/douyin_6571001202379590925.mp4'

# # 定义输出音频文件的路径
# audio_file_path = '/mnt/data10t/dazuoye/GROUP2024-GEN6/FakeSV/dataset/douyin_6571001202379590925.mp3'

# # 加载视频文件
# video_clip = VideoFileClip(video_file_path)

# # 提取音频
# audio_clip = video_clip.audio

# # 将音频保存为文件
# audio_clip.write_audiofile(audio_file_path)

# # 关闭VideoClip对象以释放资源
# video_clip.close()
# audio_clip.close()

import os
import torch
import numpy as np
from moviepy.editor import VideoFileClip
import pickle

from vggish_modified import VGGish as VGGish_modified  # 修改后的VGGish代码

# 定义输入视频文件夹和输出特征文件路径
video_file_path = '/mnt/data10t/dazuoye/GROUP2024-GEN6/FakeSV/dataset/videos/douyin_6571001202379590925.mp4'
feature_path = '/mnt/data10t/dazuoye/GROUP2024-GEN6/FakeSV/code/VGGish_Feature_Extractor/my_dict_vid_audioconvfea.pkl'

# 加载模型
urls = {
    'vggish': 'https://github.com/harritaylor/torchvggish/'
              'releases/download/v0.1/vggish-10086976.pth',
    'pca': 'https://github.com/harritaylor/torchvggish/'
           'releases/download/v0.1/vggish_pca_params-970ea276.pth'
}
vggish_model = VGGish_modified(urls, pretrained=True)

# 初始化保存特征的字典
features_dict = {}

audio_file_path = '/mnt/data10t/dazuoye/GROUP2024-GEN6/FakeSV/dataset/1.wav'


# 提取视频文件名（不包括扩展名）作为视频ID
video_file_name = os.path.basename(video_file_path)
video_id = os.path.splitext(video_file_name)[0]
# video_id = video_file_name.split('_')[1].split('.')[0]

# 从视频中提取音频
video = VideoFileClip(video_file_path)
video.audio.write_audiofile(audio_file_path)

# 提取特征
features = vggish_model(audio_file_path)


# 保存特征到字典中
features_dict[video_id] = features.cpu().detach().numpy()


# 保存特征字典到文件
with open(feature_path, 'wb') as f:
    pickle.dump(features_dict, f)

print(f"Audio features have been saved to {feature_path}")