Datasets:
0x22almostEvil
commited on
Commit
•
7beabae
1
Parent(s):
9d051db
Upload 5 files
Browse files- channel_urls.txt +1 -0
- generate_dataset.py +76 -0
- generate_urls.py +10 -0
- polish_dataset.py +33 -0
- urls.txt +0 -0
channel_urls.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
https://www.youtube.com/channel/UCX6OQ3DkcsbYNE6H8uQQuVA
|
generate_dataset.py
ADDED
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import subprocess
|
2 |
+
import os
|
3 |
+
import glob
|
4 |
+
import shutil
|
5 |
+
from concurrent.futures import ThreadPoolExecutor
|
6 |
+
|
7 |
+
def move_to_unmatched(base_path, lang, audio_exists, subtitle_exists):
|
8 |
+
unmatched_path = os.path.join(base_path, "unmatched", lang)
|
9 |
+
os.makedirs(unmatched_path, exist_ok=True)
|
10 |
+
|
11 |
+
if audio_exists:
|
12 |
+
audio_file = glob.glob(f"{base_path}/{lang}.wav")[0]
|
13 |
+
shutil.move(audio_file, unmatched_path)
|
14 |
+
|
15 |
+
if subtitle_exists:
|
16 |
+
subtitle_file = glob.glob(f"{base_path}/{lang}.vtt")[0]
|
17 |
+
shutil.move(subtitle_file, unmatched_path)
|
18 |
+
|
19 |
+
def download_and_convert_audio(video_url, languages):
|
20 |
+
video_id = video_url.split('=')[-1]
|
21 |
+
base_path = f"dataset/{video_id}"
|
22 |
+
os.makedirs(base_path, exist_ok=True)
|
23 |
+
|
24 |
+
for lang in languages:
|
25 |
+
command_download = [
|
26 |
+
'yt-dlp',
|
27 |
+
'--extract-audio',
|
28 |
+
'--audio-format', 'm4a',
|
29 |
+
'--write-sub',
|
30 |
+
'--sub-langs', lang,
|
31 |
+
'--sub-format', 'vtt',
|
32 |
+
'--output', f"{base_path}/%(title)s.{lang}.%(ext)s",
|
33 |
+
'--format', f'bestaudio[ext=m4a][acodec=mp4a.40.2][language={lang}]',
|
34 |
+
video_url
|
35 |
+
]
|
36 |
+
subprocess.run(command_download)
|
37 |
+
|
38 |
+
m4a_files = glob.glob(f"{base_path}/*.{lang}.m4a")
|
39 |
+
if m4a_files:
|
40 |
+
m4a_file = m4a_files[0]
|
41 |
+
wav_file = f"{base_path}/{lang}.wav"
|
42 |
+
command_convert = ['ffmpeg', '-i', m4a_file, '-acodec', 'pcm_s16le', '-ar', '44100', wav_file]
|
43 |
+
subprocess.run(command_convert)
|
44 |
+
os.remove(m4a_file)
|
45 |
+
|
46 |
+
vtt_files = glob.glob(f"{base_path}/*.{lang}.vtt")
|
47 |
+
if vtt_files:
|
48 |
+
vtt_file = vtt_files[0]
|
49 |
+
new_vtt_file = f"{base_path}/{lang}.vtt"
|
50 |
+
os.rename(vtt_file, new_vtt_file)
|
51 |
+
|
52 |
+
audio_exists = bool(glob.glob(f"{base_path}/{lang}.wav"))
|
53 |
+
subtitle_exists = bool(glob.glob(f"{base_path}/{lang}.vtt"))
|
54 |
+
|
55 |
+
if not (audio_exists and subtitle_exists):
|
56 |
+
move_to_unmatched(base_path, lang, audio_exists, subtitle_exists)
|
57 |
+
|
58 |
+
def process_video(url, languages):
|
59 |
+
try:
|
60 |
+
download_and_convert_audio(url, languages)
|
61 |
+
except Exception as e:
|
62 |
+
print(f"Error processing {url}: {e}")
|
63 |
+
|
64 |
+
if __name__ == "__main__":
|
65 |
+
|
66 |
+
video_urls = []
|
67 |
+
|
68 |
+
with open('urls.txt', 'r') as file:
|
69 |
+
video_urls = [line.strip() for line in file if line.strip()]
|
70 |
+
|
71 |
+
languages = ["ar", "es", "fr", "hi", "id", "ja", "ko", "pt", "ru", "th", "tr", "vi", "en"]
|
72 |
+
|
73 |
+
with ThreadPoolExecutor(max_workers=5) as executor:
|
74 |
+
futures = [executor.submit(process_video, url, languages) for url in video_urls]
|
75 |
+
for future in futures:
|
76 |
+
future.result()
|
generate_urls.py
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import subprocess
|
2 |
+
|
3 |
+
video_urls = []
|
4 |
+
|
5 |
+
with open("channel_urls.txt", "r") as f:
|
6 |
+
channel_urls = f.readlines()
|
7 |
+
|
8 |
+
for channel_url in channel_urls:
|
9 |
+
command = ["yt-dlp", "--flat-playlist", "--print-to-file", "webpage_url", "urls.txt", channel_url]
|
10 |
+
subprocess.run(command)
|
polish_dataset.py
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import shutil
|
3 |
+
|
4 |
+
main_directory = "dataset"
|
5 |
+
|
6 |
+
deleted_folders = 0
|
7 |
+
remaining_folders = 0
|
8 |
+
|
9 |
+
for folder_name in os.listdir(main_directory):
|
10 |
+
folder_path = os.path.join(main_directory, folder_name)
|
11 |
+
|
12 |
+
if os.path.isdir(folder_path):
|
13 |
+
subfolders = os.listdir(folder_path)
|
14 |
+
if len(subfolders) == 1 and subfolders[0] == "unmatched" and os.path.isdir(os.path.join(folder_path, "unmatched")):
|
15 |
+
|
16 |
+
print(f"Content of folder {folder_name}:")
|
17 |
+
for item in os.listdir(folder_path):
|
18 |
+
print(f"- {item}")
|
19 |
+
|
20 |
+
print(f"It's bad sample. Deleting folder: {folder_name}")
|
21 |
+
shutil.rmtree(folder_path)
|
22 |
+
deleted_folders += 1
|
23 |
+
|
24 |
+
else:
|
25 |
+
print(f"Content of folder {folder_name}:")
|
26 |
+
for item in os.listdir(folder_path):
|
27 |
+
print(f"- {item}")
|
28 |
+
print(f"It's good sample #{folder_name}. Keeping.")
|
29 |
+
remaining_folders += 1
|
30 |
+
|
31 |
+
# Display the results
|
32 |
+
print(f"Deleted {deleted_folders} empty folders.")
|
33 |
+
print(f"Remaining folders: {remaining_folders}")
|
urls.txt
ADDED
File without changes
|