$ pip install pyannote.server
$ python -m pyannote.server.run
-
/parser/
returns list of supported file formats$ curl -X GET http://localhost:5000/parser/ ["mdtm", "uem"]
-
/parser/<format>
parsesPOST
ed file and returns its content in PyAnnote JSON format.
-
/metric/
returns list of available evaluation metrics$ curl -X GET http://localhost:5000/metric/ ["detection", "diarization", "identification"]
-
/metric/<name>
comparesPOST
ed reference and hypothesis annotations in JSON format and returns the corresponding evaluation metric.Input format (JSON)
{ "reference": [ ... ], "hypothesis": [ ... ] }
Output format (JSON)
{ METRIC: { METRIC: value, COMPONENT_1: value_1, COMPONENT_2: value_2, ... # components are values from ... # which the final value is computed }, ... # one call to /parser/<metric> may ... # return more than one sub-metrics }
-
/error/diff
comparesPOST
ed reference and hypothesis and returns their differences.Input format (JSON)
# same format as for metric/<name> { "reference": [ ... ], "hypothesis": [ ... ] }
-
/error/regression
comparesPOST
ed reference with two hypotheses and returns regressions and/or improvements brought by the second one (after
) over the first one (before
).Input format (JSON)
{ "reference": [ ... ], "before": [ ... ], "after": [ ... ] }