path = OCR.format(repository=REPOSITORY, uri=uri)
names = ['start', 'end', 'start_frame', 'end_frame', 'name', 'confidence']
pyannote_ocr = Annotation(uri=uri)
try:
ocr = pd.read_table(path, delim_whitespace=True, header=None, names=names)
for _, (start, end, _, _, name, _) in ocr.iterrows():
pyannote_ocr[Segment(start, end)] = name
except pandas.parser.CParserError as e:
pass
# name each speaker by most co-occurring OCR name
if not pyannote_ocr:
named_speakers = Annotation(uri=uri)
else:
named_speakers = argmax_tagger(pyannote_ocr, pyannote_speakers)
named_speakers = named_speakers.subset(pyannote_ocr.labels())
path = FUSION.format(repository=REPOSITORY, uri=uri)
directory = os.path.dirname(path)
if not os.path.exists(directory):
os.makedirs(directory)
with open(path, 'w') as fp:
duplicates = dict()
for (speech_turn, track), (_, shot_id) in named_speakers.co_iter(pyannote_shots):
original_person_name = named_speakers[speech_turn, track]
person_name = mapping.setdefault(
