khuangaf

def second2timecode(time):
    '''Convert second into time code for srt.'''
    hours = time // 3600
    time -= 3600 * hours
    minutes = time // 60
    time -= 60 * minutes
    secs = np.floor(time % 60)
    msecs = (time % 60 - secs) * 1000
    timecode = "%02d:%02d:%02d,%03d" % (hours, minutes, secs, msecs)
    return timecode

khuangaf

def same_subtitle(current_subtitle, next_subtitle):
    '''Return true if the two given subtitles are the same (but can tolerate a bit difference)'''
    # convert the two subtitle into set e.g. '我很乖' -> {'我','很','乖'}
    current_set = set(current_subtitle)
    next_set = set(next_subtitle)
    current_set_len = len(current_set)
    next_set_len = len(next_set)
    intersect_set = current_set & next_set
    intersect_set_len = len(intersect_set)
    

khuangaf

df = pd.DataFrame(columns=['id', 'prediction', 'confidence'])
df['id'] = image_ids
df['prediction'] = predictions
df['confidence'] = confidences
df.to_csv(f'{args.results_dir}/{video_id}.csv', index=None)

khuangaf

# the endpoint of the api
ENDPOINT_URL = 'https://vision.googleapis.com/v1p1beta1/images:annotate'
# the api key
os.environ['GOOGLE_APPLICATION_CREDENTIALS']='your/path/to/credentials.json'
# init api client
client = vision.ImageAnnotatorClient()

with io.open(YOUR_IMAGE_PATH, 'rb') as f:
        content = f.read()
image = vision.types.Image(content=content)    

khuangaf

class SubtitleConfig(Config):
    """Configuration for training on the nucleus segmentation dataset."""
    # Give the configuration a recognizable name
    NAME = "subtitle"

    # Adjust depending on your GPU memory
    IMAGES_PER_GPU = 4

    # Number of classes (including background)
    NUM_CLASSES = 1 + 1  # Background + subtitle

khuangaf

def get_multi_masks(img):
    '''
    Get connected components (multi-masks)
    args: 
        img: np.ndarray
    return:
        np.ndarray with same dimension as the input
    '''

    # Threshold the image to make it become either 0 or 1

khuangaf

def video2frames(video_id, video_path, processed_videos_dir, frame_path, sample_rate):
    """
    Execute shell command which calls ffmpeg to split video into frames.
    Parameters
    ----------
    video_id : str
        The video id of a video
    video_path: str
        The directory storing the videos
    processed_video_path: str

khuangaf

def extract_audio(video_id,videos_dir, audios_dir):
    """
    Download the videos
    Parameters
    ----------
    video_id : str
        A Youtube video id. 
    videos_dir: str
        The directory which stores videos.
    audios_dir: str

khuangaf

def download_video(video_id, path="videos", verbose=True):
    
    """
    Download the videos
    Parameters
    ----------
    video_id : str
        A Youtube video id. 
    path: str
        The directory which stores videos.

khuangaf

def get_video_ids(playlist_id):
    """
    Get the video ids given a playlist id.
    Parameters
    ----------
    playlist_id : str
        A Youtube playlist id. (up to 50 results)
    Returns
    -------
    video_ids: list(str)