def overlapping_iter(self, t):
"""Like `overlapping` but returns a segment iterator instead
See also
--------
:func:`pyannote.core.Timeline.overlapping`
"""
segment = Segment(start=t, end=t)
iterable = self.segments_list_.irange(maximum=segment)
for segment in self.segments_list_.irange(maximum=segment):
if segment.overlaps(t):
yield segment
def co_iter(self, other):
"""Iterate over pairs of intersecting segments
>>> timeline1 = Timeline([Segment(0, 2), Segment(1, 2), Segment(3, 4)])
>>> timeline2 = Timeline([Segment(1, 3), Segment(3, 5)])
>>> for segment1, segment2 in timeline1.co_iter(timeline2):
... print(segment1, segment2)
(<Segment(0, 2)>, <Segment(1, 3)>)
(<Segment(1, 2)>, <Segment(1, 3)>)
(<Segment(3, 4)>, <Segment(3, 5)>)
Parameters
----------
other : Timeline
Second timeline
Returns
-------
iterable : (Segment, Segment) iterable
Yields pairs of intersecting segments in chronological order.
"""
for segment in self.segments_list_:
# iterate over segments that starts before 'segment' ends
temp = Segment(start=segment.end, end=segment.end)
for other_segment in other.segments_list_.irange(maximum=temp):
if segment.intersects(other_segment):
yield segment, other_segment
def gaps_iter(self, support=None):
"""Like `gaps` but returns a segment iterator instead
See also
--------
:func:`pyannote.core.Timeline.gaps`
"""
if support is None:
support = self.extent()
if not isinstance(support, (Segment, Timeline)):
raise TypeError("unsupported operand type(s) for -':"
"%s and Timeline." % type(support).__name__)
# segment support
if isinstance(support, Segment):
# `end` is meant to store the end time of former segment
# initialize it with beginning of provided segment `support`
end = support.start
# support on the intersection of timeline and provided segment
for segment in self.crop(support, mode='intersection').support():
# add gap between each pair of consecutive segments
# if there is no gap, segment is empty, therefore not added
gap = Segment(start=end, end=segment.start)
if gap:
yield gap
# keep track of the end of former segment
end = segment.end
# add final gap (if not empty)
gap = Segment(start=end, end=support.end)
if gap:
yield gap
# timeline support
elif isinstance(support, Timeline):
# yield gaps for every segment in support of provided timeline
for segment in support.support():
for gap in self.gaps_iter(support=segment):
yield gap
def apply(self, predictions, dimension=0):
"""Peak detection
Parameter
---------
predictions : SlidingWindowFeature
Predictions returned by segmentation approaches.
Returns
-------
segmentation : Timeline
Partition.
"""
if len(predictions.data.shape) == 1:
y = predictions.data
elif predictions.data.shape[1] == 1:
y = predictions.data[:, 0]
else:
y = predictions.data[:, dimension]
if self.log_scale:
y = np.exp(y)
sw = predictions.sliding_window
precision = sw.step
order = max(1, int(np.rint(self.min_duration / precision)))
indices = scipy.signal.argrelmax(y, order=order)[0]
if self.scale == 'absolute':
mini = 0
maxi = 1
elif self.scale == 'relative':
mini = np.nanmin(y)
maxi = np.nanmax(y)
elif self.scale == 'percentile':
mini = np.nanpercentile(y, 1)
maxi = np.nanpercentile(y, 99)
threshold = mini + self.alpha * (maxi - mini)
peak_time = np.array(
[sw[i].middle for i in indices if y[i] > threshold])
n_windows = len(y)
start_time = sw[0].start
end_time = sw[n_windows].end
boundaries = np.hstack([[start_time], peak_time, [end_time]])
segmentation = Timeline()
for i, (start, end) in enumerate(pairwise(boundaries)):
segment = Segment(start, end)
segmentation.add(segment)
return segmentation
def segmentation(self):
"""Segmentation
Create the unique timeline with same support and same set of segment
boundaries as original timeline, but with no overlapping segments.
A picture is worth a thousand words::
timeline
|------| |------| |----|
|--| |-----| |----------|
timeline.segmentation()
|-|--|-| |-|---|--| |--|----|--|
Returns
-------
timeline : Timeline
(unique) timeline with same support and same set of segment
boundaries as original timeline, but with no overlapping segments.
"""
# COMPLEXITY: O(n)
support = self.support()
# COMPLEXITY: O(n.log n)
# get all boundaries (sorted)
# |------| |------| |----|
# |--| |-----| |----------|
# becomes
# | | | | | | | | | | | |
timestamps = set([])
for (start, end) in self:
timestamps.add(start)
timestamps.add(end)
timestamps = sorted(timestamps)
# create new partition timeline
# | | | | | | | | | | | |
# becomes
# |-|--|-| |-|---|--| |--|----|--|
# start with an empty copy
timeline = Timeline(uri=self.uri)
if len(timestamps) == 0:
return Timeline(uri=self.uri)
segments = []
start = timestamps[0]
for end in timestamps[1:]:
# only add segments that are covered by original timeline
segment = Segment(start=start, end=end)
if segment and support.overlapping(segment.middle):
segments.append(segment)
# next segment...
start = end
return Timeline(segments=segments, uri=self.uri)
def from_json(cls, data):
"""Deserialization
See also
--------
:mod:`pyannote.core.json`
"""
uri = data.get(PYANNOTE_URI, None)
segments = [Segment.from_json(s) for s in data[PYANNOTE_JSON_CONTENT]]
return cls(segments=segments, uri=uri)
def extent(self):
"""Extent
The extent of a timeline is the segment of minimum duration that
contains every segments of the timeline. It is unique, by definition.
The extent of an empty timeline is an empty segment.
A picture is worth a thousand words::
timeline
|------| |------| |----|
|--| |-----| |----------|
timeline.extent()
|--------------------------------|
Returns
-------
extent : Segment
Timeline extent
Examples
--------
>>> timeline = Timeline(segments=[Segment(0, 1), Segment(9, 10)])
>>> timeline.extent()
<Segment(0, 10)>
"""
if self.segments_set_:
segments_boundaries_ = self.segments_boundaries_
start = segments_boundaries_[0]
end = segments_boundaries_[-1]
return Segment(start=start, end=end)
else:
import numpy as np
return Segment(start=np.inf, end=-np.inf)
def apply(self, predictions, dimension=0):
"""
Parameters
----------
predictions : SlidingWindowFeature
Must be mono-dimensional
dimension : int, optional
Which dimension to process
"""
if len(predictions.data.shape) == 1:
data = predictions.data
elif predictions.data.shape[1] == 1:
data = predictions.data[:, 0]
else:
data = predictions.data[:, dimension]
if self.log_scale:
data = np.exp(data)
n_samples = predictions.getNumber()
window = predictions.sliding_window
timestamps = [window[i].middle for i in range(n_samples)]
# initial state
start = timestamps[0]
label = data[0] > self.onset
if self.scale == 'absolute':
mini = 0
maxi = 1
elif self.scale == 'relative':
mini = np.nanmin(data)
maxi = np.nanmax(data)
elif self.scale == 'percentile':
mini = np.nanpercentile(data, 1)
maxi = np.nanpercentile(data, 99)
onset = mini + self.onset * (maxi - mini)
offset = mini + self.offset * (maxi - mini)
# timeline meant to store 'active' segments
active = Timeline()
for t, y in zip(timestamps[1:], data[1:]):
# currently active
if label:
# switching from active to inactive
if y < offset:
segment = Segment(start - self.pad_onset,
t + self.pad_offset)
active.add(segment)
start = t
label = False
# currently inactive
else:
# switching from inactive to active
if y > onset:
start = t
label = True
# if active at the end, add final segment
if label:
segment = Segment(start - self.pad_onset, t + self.pad_offset)
active.add(segment)
# because of padding, some 'active' segments might be overlapping
# therefore, we merge those overlapping segments
active = active.support()
# remove short 'active' segments
active = Timeline(
[s for s in active if s.duration > self.min_duration_on])
# fill short 'inactive' segments
inactive = active.gaps()
for s in inactive:
if s.duration < self.min_duration_off:
active.add(s)
active = active.support()
return active