def wrap_histogram(hist_, edges_, DEBUG_ROTINVAR=False):
r"""
Simulates the first and last histogram bin being being adjacent to one another
by replicating those bins at the last and first positions respectively.
Args:
hist_ (ndarray):
edges_ (ndarray):
Returns:
tuple: (hist_wrap, edge_wrap)
CommandLine:
python -m vtool.histogram --test-wrap_histogram
Example:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> hist_ = np.array([ 8. , 0. , 0. , 34.32, 29.45, 0. , 0. , 6.73])
>>> edges_ = np.array([ 0. , 0.78539816, 1.57079633,
... 2.35619449, 3.14159265, 3.92699081,
... 4.71238898, 5.49778714, 6.2831853 ])
>>> # execute function
>>> (hist_wrap, edge_wrap) = wrap_histogram(hist_, edges_)
>>> # verify results
>>> edgewrap_str = '[' + ', '.join(['%.2f' % _ for _ in edge_wrap]) + ']'
>>> histwrap_str = str(hist_wrap.tolist())
>>> result = histwrap_str + ut.NEWLINE + edgewrap_str
>>> print(result)
[6.73, 8.0, 0.0, 0.0, 34.32, 29.45, 0.0, 0.0, 6.73, 8.0]
[-0.79, 0.00, 0.79, 1.57, 2.36, 3.14, 3.93, 4.71, 5.50, 6.28, 7.07]
"""
# FIXME; THIS NEEDS INFORMATION ABOUT THE DISTANCE FROM THE LAST BIN
# TO THE FIRST. IT IS OK AS LONG AS ALL STEPS ARE EQUAL, BUT IT IS NOT
# GENERAL
left_step, right_step = np.diff(edges_)[[0, -1]]
hist_wrap = np.hstack((hist_[-1:], hist_, hist_[0:1]))
edge_wrap = np.hstack(
(edges_[0:1] - left_step, edges_, edges_[-1:] + right_step))
if DEBUG_ROTINVAR:
import vtool as vt
print(vt.kpts_docrepr(hist_wrap, 'hist_wrap', False))
print(vt.kpts_docrepr(edge_wrap, 'edge_wrap', False))
return hist_wrap, edge_wrap
def wrap_histogram(hist_, edges_, DEBUG_ROTINVAR=False):
r"""
Simulates the first and last histogram bin being being adjacent to one another
by replicating those bins at the last and first positions respectively.
Args:
hist_ (ndarray):
edges_ (ndarray):
Returns:
tuple: (hist_wrap, edge_wrap)
CommandLine:
python -m vtool.histogram --test-wrap_histogram
Example:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> hist_ = np.array([ 8. , 0. , 0. , 34.32, 29.45, 0. , 0. , 6.73])
>>> edges_ = np.array([ 0. , 0.78539816, 1.57079633,
... 2.35619449, 3.14159265, 3.92699081,
... 4.71238898, 5.49778714, 6.2831853 ])
>>> # execute function
>>> (hist_wrap, edge_wrap) = wrap_histogram(hist_, edges_)
>>> # verify results
>>> edgewrap_str = '[' + ', '.join(['%.2f' % _ for _ in edge_wrap]) + ']'
>>> histwrap_str = str(hist_wrap.tolist())
>>> result = histwrap_str + ut.NEWLINE + edgewrap_str
>>> print(result)
[6.73, 8.0, 0.0, 0.0, 34.32, 29.45, 0.0, 0.0, 6.73, 8.0]
[-0.79, 0.00, 0.79, 1.57, 2.36, 3.14, 3.93, 4.71, 5.50, 6.28, 7.07]
"""
# FIXME; THIS NEEDS INFORMATION ABOUT THE DISTANCE FROM THE LAST BIN
# TO THE FIRST. IT IS OK AS LONG AS ALL STEPS ARE EQUAL, BUT IT IS NOT
# GENERAL
left_step, right_step = np.diff(edges_)[[0, -1]]
hist_wrap = np.hstack((hist_[-1:], hist_, hist_[0:1]))
edge_wrap = np.hstack((edges_[0:1] - left_step, edges_, edges_[-1:] + right_step))
if DEBUG_ROTINVAR:
import vtool as vt
print(vt.kpts_docrepr(hist_wrap, 'hist_wrap', False))
print(vt.kpts_docrepr(edge_wrap, 'edge_wrap', False))
return hist_wrap, edge_wrap
def interpolated_histogram(data, weights, range_, bins, interpolation_wrap=True,
DEBUG_ROTINVAR=False):
r"""
Follows np.histogram, but does interpolation
Args:
range_ (tuple): range from 0 to 1
bins (?):
CommandLine:
python -m vtool.histogram --test-interpolated_histogram
Example0:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> data = np.array([ 0, 1, 2, 3.5, 3, 3, 4, 4])
>>> weights = np.array([1., 1., 1., 1., 1., 1., 1., 1.])
>>> range_ = (0, 4)
>>> bins = 5
>>> interpolation_wrap = False
>>> # execute function
>>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
>>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
>>> assert hist.size == bins
>>> assert edges.size == bins + 1
>>> result = get_histinfo_str(hist, edges)
>>> print(result)
Example1:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> data = np.array([ 0, 1, 2, 3.5, 3, 3, 4, 4])
>>> weights = np.array([4.5, 1., 1., 1., 1., 1., 1., 1.])
>>> range_ = (-.5, 4.5)
>>> bins = 5
>>> interpolation_wrap = True
>>> # execute function
>>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
>>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
>>> assert hist.size == bins
>>> assert edges.size == bins + 1
>>> result = get_histinfo_str(hist, edges)
>>> print(result)
#Example2:
# >>> # ENABLE_DOCTEST
# >>> from vtool.histogram import * # NOQA
# >>> # build test data
# >>> data = np.random.rand(10)
# >>> weights = np.random.rand(10)
# >>> range_ = (0, 1)
# >>> bins = np.random.randint(2) + 1 + np.random.randint(2) * 100
# >>> interpolation_wrap = True
# >>> # execute function
# >>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
# >>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
# >>> assert hist.size == bins
# >>> assert edges.size == bins + 1
# >>> result = get_histinfo_str(hist, edges)
# >>> print(result)
"""
assert bins > 0, 'must have nonzero bins'
data = np.asarray(data)
if weights is not None:
weights = np.asarray(weights)
assert np.all(weights.shape == data.shape), 'shapes disagree'
weights = weights.ravel()
data = data.ravel()
# Compute bin edges like in np.histogram
start, stop = float(range_[0]), float(range_[1])
if start == stop:
start -= 0.5
stop += 0.5
# Find bin edges
hist_dtype = np.float64
# Compute bin step size, add one if last bin is the same as the first
step = (stop - start) / float((bins + interpolation_wrap))
#edges = [start + i * step for i in range(bins + 1)]
#centers = hist_edges_to_centers(edges)
half_step = step / 2.0
# Find fractional bin center index for each datapoint
data_offset = start + half_step
frac_index = (data - data_offset) / step
# Find bin center to the left of each datapoint
left_index = np.floor(frac_index).astype(np.int32)
# Find bin center to the right of each datapoint
right_index = left_index + 1
# Find the fraction of the distiance the right center is away from the datapoint
right_alpha = (frac_index - left_index)
left_alpha = 1.0 - right_alpha
if DEBUG_ROTINVAR:
print('bins = %r' % bins)
print('step = %r' % step)
print('half_step = %r' % half_step)
print('data_offset = %r' % data_offset)
TAU = 2 * np.pi
print("-.5 MOD tau = %r" % (-.5 % TAU,))
# Handle edge cases
if interpolation_wrap:
# when the stop == start (like in orientations)
left_index %= bins
right_index %= bins
else:
left_index[left_index < 0] = 0
right_index[right_index >= bins] = bins - 1
# Each keypoint votes into its left and right bins
left_vote = left_alpha * weights
right_vote = right_alpha * weights
hist = np.zeros((bins,), hist_dtype)
# TODO: can problably do this faster with cumsum
for index, vote in zip(left_index, left_vote):
hist[index] += vote
for index, vote in zip(right_index, right_vote):
hist[index] += vote
if interpolation_wrap:
edges = np.linspace(start, stop, bins + 1, endpoint=False)
else:
edges = np.linspace(start, stop, bins + 1, endpoint=True)
if DEBUG_ROTINVAR:
import vtool as vt
assert np.allclose(np.diff(edges), step)
print(hist.shape)
print(edges.shape)
print(vt.kpts_docrepr(hist, 'hist', False))
print(vt.kpts_docrepr(edges, 'edges', False))
return hist, edges
def interpolated_histogram(data,
weights,
range_,
bins,
interpolation_wrap=True,
DEBUG_ROTINVAR=False):
r"""
Follows np.histogram, but does interpolation
Args:
range_ (tuple): range from 0 to 1
bins (?):
CommandLine:
python -m vtool.histogram --test-interpolated_histogram
Example0:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> data = np.array([ 0, 1, 2, 3.5, 3, 3, 4, 4])
>>> weights = np.array([1., 1., 1., 1., 1., 1., 1., 1.])
>>> range_ = (0, 4)
>>> bins = 5
>>> interpolation_wrap = False
>>> # execute function
>>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
>>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
>>> assert hist.size == bins
>>> assert edges.size == bins + 1
>>> result = get_histinfo_str(hist, edges)
>>> print(result)
Example1:
>>> # ENABLE_DOCTEST
>>> from vtool.histogram import * # NOQA
>>> # build test data
>>> data = np.array([ 0, 1, 2, 3.5, 3, 3, 4, 4])
>>> weights = np.array([4.5, 1., 1., 1., 1., 1., 1., 1.])
>>> range_ = (-.5, 4.5)
>>> bins = 5
>>> interpolation_wrap = True
>>> # execute function
>>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
>>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
>>> assert hist.size == bins
>>> assert edges.size == bins + 1
>>> result = get_histinfo_str(hist, edges)
>>> print(result)
#Example2:
# >>> # ENABLE_DOCTEST
# >>> from vtool.histogram import * # NOQA
# >>> # build test data
# >>> data = np.random.rand(10)
# >>> weights = np.random.rand(10)
# >>> range_ = (0, 1)
# >>> bins = np.random.randint(2) + 1 + np.random.randint(2) * 100
# >>> interpolation_wrap = True
# >>> # execute function
# >>> hist, edges = interpolated_histogram(data, weights, range_, bins, interpolation_wrap)
# >>> assert np.abs(hist.sum() - weights.sum()) < 1E-9
# >>> assert hist.size == bins
# >>> assert edges.size == bins + 1
# >>> result = get_histinfo_str(hist, edges)
# >>> print(result)
"""
assert bins > 0, 'must have nonzero bins'
data = np.asarray(data)
if weights is not None:
weights = np.asarray(weights)
assert np.all(weights.shape == data.shape), 'shapes disagree'
weights = weights.ravel()
data = data.ravel()
# Compute bin edges like in np.histogram
start, stop = float(range_[0]), float(range_[1])
if start == stop:
start -= 0.5
stop += 0.5
# Find bin edges
hist_dtype = np.float64
# Compute bin step size, add one if last bin is the same as the first
step = (stop - start) / float((bins + interpolation_wrap))
#edges = [start + i * step for i in range(bins + 1)]
#centers = hist_edges_to_centers(edges)
half_step = step / 2.0
# Find fractional bin center index for each datapoint
data_offset = start + half_step
frac_index = (data - data_offset) / step
# Find bin center to the left of each datapoint
left_index = np.floor(frac_index).astype(np.int32)
# Find bin center to the right of each datapoint
right_index = left_index + 1
# Find the fraction of the distiance the right center is away from the datapoint
right_alpha = (frac_index - left_index)
left_alpha = 1.0 - right_alpha
if DEBUG_ROTINVAR:
print('bins = %r' % bins)
print('step = %r' % step)
print('half_step = %r' % half_step)
print('data_offset = %r' % data_offset)
TAU = 2 * np.pi
print("-.5 MOD tau = %r" % (-.5 % TAU, ))
# Handle edge cases
if interpolation_wrap:
# when the stop == start (like in orientations)
left_index %= bins
right_index %= bins
else:
left_index[left_index < 0] = 0
right_index[right_index >= bins] = bins - 1
# Each keypoint votes into its left and right bins
left_vote = left_alpha * weights
right_vote = right_alpha * weights
hist = np.zeros((bins, ), hist_dtype)
# TODO: can problably do this faster with cumsum
for index, vote in zip(left_index, left_vote):
hist[index] += vote
for index, vote in zip(right_index, right_vote):
hist[index] += vote
if interpolation_wrap:
edges = np.linspace(start, stop, bins + 1, endpoint=False)
else:
edges = np.linspace(start, stop, bins + 1, endpoint=True)
if DEBUG_ROTINVAR:
import vtool as vt
assert np.allclose(np.diff(edges), step)
print(hist.shape)
print(edges.shape)
print(vt.kpts_docrepr(hist, 'hist', False))
print(vt.kpts_docrepr(edges, 'edges', False))
return hist, edges