def add_parser_group(self, d, target_obj):
gp_nm = d.pop('name')
if 'title' not in d:
d['title'] = gp_nm
args = util.coerce_to_iter(d.pop('arguments', None))
gp_kwargs = util.filter_kwargs(d, argparse._ArgumentGroup.__init__)
gp_obj = target_obj.add_argument_group(**gp_kwargs)
self.parser_groups[gp_nm] = gp_obj
for arg in args:
self.add_parser_argument(arg, gp_obj, gp_nm)
def add_parser_group(self, d, target_obj):
gp_nm = d.pop('name')
_ = d.setdefault('title', gp_nm)
args = util.coerce_to_iter(d.pop('arguments', None))
if args:
# only add group if it has > 0 arguments
gp_kwargs = util.filter_kwargs(d, argparse._ArgumentGroup.__init__)
gp_obj = target_obj.add_argument_group(**gp_kwargs)
self.parser_groups[gp_nm] = gp_obj
for arg in args:
self.add_parser_argument(arg, gp_obj, gp_nm)
def __init__(self, impl='KMEANS', **kwargs):
# Finally, we create the model and let 'r rip!
class_name = '_{}'.format(impl)
model_class = getattr(sys.modules[__name__], class_name)
kwargs, unused, _ = util.filter_kwargs(model_class, kwargs)
if len(unused) > 0:
logger.warning(
'Cluster model "{}" has unused arguments: {}'.format(
impl,
unused
)
)
self._model = model_class(**kwargs)
def make_parser(self, d):
args = util.coerce_to_iter(d.pop('arguments', None))
arg_groups = util.coerce_to_iter(d.pop('argument_groups', None))
d['formatter_class'] = CustomHelpFormatter
p_kwargs = util.filter_kwargs(d, argparse.ArgumentParser.__init__)
p = argparse.ArgumentParser(**p_kwargs)
for arg in args:
# add arguments not in any group
self.add_parser_argument(arg, p, 'parser')
for group in arg_groups:
# add groups and arguments therein
self.add_parser_group(group, p)
return p
def __init__(self, impl='MalletLda', **kwargs):
# Finally, we create the topic model and let 'r rip!
class_name = '_{}'.format(impl)
model_class = getattr(sys.modules[__name__], class_name)
kwargs, unused, _ = util.filter_kwargs(model_class, kwargs)
if len(unused) > 0:
logger.warning('Topic model "{}" has unused arguments: {}'.format(
impl,
unused
))
regen = kwargs.get('regen', None)
if regen:
logger.info('Forcing regeneration of topic model')
self.model = model_class(**kwargs)
def evaluate(reference_sources, estimated_sources, **kwargs):
"""Compute all metrics for the given reference and estimated annotations.
Examples
--------
>>> # reference_sources[n] should be an ndarray of samples of the
>>> # n'th reference source
>>> # estimated_sources[n] should be the same for the n'th estimated
>>> scores = mir_eval.separation.evaluate(reference_sources,
... estimated_sources)
Parameters
----------
reference_sources : np.ndarray, shape=(nsrc, nsampl)
matrix containing true sources
estimated_sources : np.ndarray, shape=(nsrc, nsampl)
matrix containing estimated sources
kwargs
Additional keyword arguments which will be passed to the
appropriate metric or preprocessing functions.
Returns
-------
scores : dict
Dictionary of scores, where the key is the metric name (str) and
the value is the (float) score achieved.
"""
# Compute all the metrics
scores = collections.OrderedDict()
sdr, sir, sar, perm = util.filter_kwargs(bss_eval_sources,
reference_sources,
estimated_sources, **kwargs)
scores['Source to Distortion'] = sdr.tolist()
scores['Source to Interference'] = sir.tolist()
scores['Source to Artifact'] = sar.tolist()
scores['Source permutation'] = perm
return scores
def evaluate(reference_sources, estimated_sources, **kwargs):
"""Compute all metrics for the given reference and estimated annotations.
Examples
--------
>>> # reference_sources[n] should be an ndarray of samples of the
>>> # n'th reference source
>>> # estimated_sources[n] should be the same for the n'th estimated
>>> scores = mir_eval.separation.evaluate(reference_sources,
... estimated_sources)
Parameters
----------
reference_sources : np.ndarray, shape=(nsrc, nsampl)
matrix containing true sources
estimated_sources : np.ndarray, shape=(nsrc, nsampl)
matrix containing estimated sources
kwargs
Additional keyword arguments which will be passed to the
appropriate metric or preprocessing functions.
Returns
-------
scores : dict
Dictionary of scores, where the key is the metric name (str) and
the value is the (float) score achieved.
"""
# Compute all the metrics
scores = collections.OrderedDict()
sdr, sir, sar, perm = util.filter_kwargs(bss_eval_sources,
reference_sources,
estimated_sources, **kwargs)
scores['Source to Distortion'] = sdr.tolist()
scores['Source to Interference'] = sir.tolist()
scores['Source to Artifact'] = sar.tolist()
scores['Source permutation'] = perm
return scores
def evaluate(reference_sources, estimated_sources, **kwargs):
"""Compute all metrics for the given reference and estimated signals.
NOTE: This will always compute :func:`mir_eval.separation.bss_eval_images`
for any valid input and will additionally compute
:func:`mir_eval.separation.bss_eval_sources` for valid input with fewer
than 3 dimensions.
Examples
--------
>>> # reference_sources[n] should be an ndarray of samples of the
>>> # n'th reference source
>>> # estimated_sources[n] should be the same for the n'th estimated source
>>> scores = mir_eval.separation.evaluate(reference_sources,
... estimated_sources)
Parameters
----------
reference_sources : np.ndarray, shape=(nsrc, nsampl[, nchan])
matrix containing true sources
estimated_sources : np.ndarray, shape=(nsrc, nsampl[, nchan])
matrix containing estimated sources
kwargs
Additional keyword arguments which will be passed to the
appropriate metric or preprocessing functions.
Returns
-------
scores : dict
Dictionary of scores, where the key is the metric name (str) and
the value is the (float) score achieved.
"""
# Compute all the metrics
scores = collections.OrderedDict()
sdr, isr, sir, sar, perm = util.filter_kwargs(bss_eval_images,
reference_sources,
estimated_sources, **kwargs)
scores['Images - Source to Distortion'] = sdr.tolist()
scores['Images - Image to Spatial'] = isr.tolist()
scores['Images - Source to Interference'] = sir.tolist()
scores['Images - Source to Artifact'] = sar.tolist()
scores['Images - Source permutation'] = perm.tolist()
sdr, isr, sir, sar, perm = util.filter_kwargs(bss_eval_images_framewise,
reference_sources,
estimated_sources, **kwargs)
scores['Images Frames - Source to Distortion'] = sdr.tolist()
scores['Images Frames - Image to Spatial'] = isr.tolist()
scores['Images Frames - Source to Interference'] = sir.tolist()
scores['Images Frames - Source to Artifact'] = sar.tolist()
scores['Images Frames - Source permutation'] = perm.tolist()
# Verify we can compute sources on this input
if reference_sources.ndim < 3 and estimated_sources.ndim < 3:
sdr, sir, sar, perm = util.filter_kwargs(bss_eval_sources_framewise,
reference_sources,
estimated_sources, **kwargs)
scores['Sources Frames - Source to Distortion'] = sdr.tolist()
scores['Sources Frames - Source to Interference'] = sir.tolist()
scores['Sources Frames - Source to Artifact'] = sar.tolist()
scores['Sources Frames - Source permutation'] = perm.tolist()
sdr, sir, sar, perm = util.filter_kwargs(bss_eval_sources,
reference_sources,
estimated_sources, **kwargs)
scores['Sources - Source to Distortion'] = sdr.tolist()
scores['Sources - Source to Interference'] = sir.tolist()
scores['Sources - Source to Artifact'] = sar.tolist()
scores['Sources - Source permutation'] = perm.tolist()
return scores