if not os.path.exists(subtitle_file) or not os.path.exists(info_file):
termcolor.cprint(
"Subtitle file or Info files do not exist. {}".format(
video_file),
color="red")
raise Exception("Subtitle file or Info files do not exist.")
# Download google subtitle to cross check with closed captions
with open(info_file) as f:
print(info_file)
print("opening json info file")
metadata = json.load(f)
print("opened json info file")
#youtube_link = metadata['webpage_url']
print("Parsing subtitle")
subtitles = load_all_subtitles(subtitle_file)
print(len(subtitles))
input = {'subtitles': subtitles, 'video_file': video_file}
overall_info["num_subtitles"] = len(subtitles)
termcolor.cprint("Got {} candidates".format(len(subtitles)),
color="yellow")
filtered_input = pipeline(input)
filtered_subtitles = filtered_input["subtitles"]
termcolor.cprint("Writing {} samples".format(len(filtered_subtitles)),
color="cyan")
for t in tqdm(filtered_subtitles):
hash = get_hash(subtitle_file + t["original_phrase"] +
str(t["ts_start"]))
wav_file_dir = os.path.join(target_dir, "wav", hash[:2])
subtitles = []
else:
overlap_ratio = [ratio(t["phrase"].lower(), s.lower())
for (t, s) in transcripts]
passed_threshold = sum(overlap_ratio) / \
len(overlap_ratio) > self.mean_wer_threshold
if not passed_threshold:
# removing all subtitles, as potentially unreliable
subtitles = []
input["subtitles"] = subtitles
return input
if __name__ == "__main__":
from youtube_helpers import load_all_subtitles
subtitles = load_all_subtitles(
"../EWmCkVfPnJ8Different_Types_of_Learning.en.vtt")
print(len(subtitles))
input = {
'subtitles': subtitles,
'video_file': ''
}
good_chars_regexp = re.compile(
r"^[A-Za-z0-9\,\.\-\?\"\'\’\!\“\s\;\:\“\”\–\‘\’\’\/\\]+$", re.IGNORECASE)
pipeline = Pipeline([
OverlappingSubtitlesRemover(),
SubtitleCaptionTextFilter(),
CaptionNormalizer(),
CaptionRegexMatcher(good_chars_regexp),
CaptionLengthFilter(min_length=5),
CaptionLeaveOnlyAlphaNumCharacters(),
SubtitleMerger(max_len_merged_sec=10),