def __axis_log(data,
n_ticks,
horiz,
sr=22050,
kwargs=None,
label='Hz',
secondary_axis='linear',
minor=None,
**_kwargs):
'''Plot a log-scaled image'''
axes_phantom = plt.gca()
if kwargs is None:
kwargs = dict()
aspect = kwargs.pop('aspect', None)
n, ticker, labeler = __get_shape_artists(data, horiz)
t_log, t_inv = __log_scale(n)
if horiz:
if minor == 'log':
ax2 = __log_scale(data.shape[0])[0]
else:
ax2 = np.linspace(0, data.shape[0], data.shape[0]).astype(int)
ax1 = t_log
else:
if minor == 'log':
ax1 = __log_scale(data.shape[1])[0]
else:
ax1 = np.linspace(0, data.shape[1], data.shape[1]).astype(int)
ax2 = t_log
args = (ax1, ax2, data)
# NOTE: 2013-11-14 16:15:33 by Brian McFee <*****@*****.**>
# We draw the image twice here. This is a hack to get around
# NonUniformImage not properly setting hooks for color.
# Drawing twice enables things like colorbar() to work properly.
im_phantom = img.NonUniformImage(axes_phantom,
extent=(args[0].min(), args[0].max(),
args[1].min(), args[1].max()),
**kwargs)
im_phantom.set_data(*args)
kwargs['aspect'] = aspect
axes_phantom.images[0] = im_phantom
positions = np.linspace(0, n - 1, n_ticks, endpoint=True).astype(int)
# One extra value here to catch nyquist
values = np.linspace(0, 0.5 * sr, n, endpoint=True).astype(int)
ticker(positions, values[t_inv[positions]])
labeler(label)
def surface_plot(self, x_range, y_range, data_matrix, plotAxis = None, interpolation_type = 'bilinear'):
#Decide axis for the image to be plotted
if plotAxis == None:
plotAxis = self.Axis
im = image.NonUniformImage(ax=plotAxis, interpolation=interpolation_type, cmap=cm.gist_earth)
#Input data
im.set_data(x_range, y_range, data_matrix)
#Plot the image
plotAxis.images.append(im)
#Set the axis limit
plotAxis.set_xlim(x_range[0],x_range[-1])
plotAxis.set_ylim(y_range[0],y_range[-1])
def specshow(data,
sr=22050,
hop_length=512,
x_axis=None,
y_axis=None,
n_xticks=5,
n_yticks=5,
fmin=None,
fmax=None,
**kwargs):
'''Display a spectrogram/chromagram/cqt/etc.
Functions as a drop-in replacement for ``matplotlib.pyplot.imshow``, but with useful defaults.
:usage:
>>> # Visualize an STFT with linear frequency scaling
>>> D = np.abs(librosa.stft(y))
>>> librosa.display.specshow(D, sr=sr, y_axis='linear')
>>> # Or with logarithmic frequency scaling
>>> librosa.display.specshow(D, sr=sr, y_axis='log')
>>> # Visualize a CQT with note markers
>>> CQT = librosa.cqt(y, sr, fmin=55, fmax=880)
>>> librosa.display.specshow(CQT, sr=sr, y_axis='cqt_note', fmin=55, fmax=880)
>>> # Draw time markers automatically
>>> librosa.display.specshow(D, sr=sr, hop_length=hop_length, x_axis='time')
>>> # Draw a chromagram with pitch classes
>>> C = librosa.feature.chromagram(y, sr)
>>> librosa.display.specshow(C, y_axis='chroma')
>>> # Force a grayscale colormap (white -> black)
>>> librosa.display.specshow(librosa.logamplitude(D), cmap='gray_r')
:parameters:
- data : np.ndarray
Matrix to display (eg, spectrogram)
- sr : int > 0
Sample rate. Used to determine time scale in x-axis
- hop_length : int > 0
Hop length. Also used to determine time scale in x-axis
- x_axis : None or {'time', 'frames', 'off'}
If None or 'off', no x axis is displayed.
If 'time', markers are shown as milliseconds, seconds, minutes, or hours.
(See ``time_ticks()`` for details.)
If 'frames', markers are shown as frame counts.
- y_axis : None or {'linear', 'mel', 'cqt_hz', 'cqt_note', 'chroma', 'off'}
- None or 'off': no y axis is displayed.
- 'linear': frequency range is determined by the FFT window and sampling rate.
- 'log': the image is displayed on a vertical log scale.
- 'mel': frequencies are determined by the mel scale.
- 'cqt_hz': frequencies are determined by the fmin and fmax values.
- 'cqt_note': pitches are determined by the fmin and fmax values.
- 'chroma': pitches are determined by the chroma filters.
- n_xticks : int > 0
If x_axis is drawn, the number of ticks to show
- n_yticks : int > 0
If y_axis is drawn, the number of ticks to show
- fmin : float > 0 or None
- fmax : float > 0 or None
Used for setting the Mel or constantq frequency scales
- kwargs
Additional keyword arguments passed through to ``matplotlib.pyplot.imshow``.
:returns:
- image : ``matplotlib.image.AxesImage``
As returned from ``matplotlib.pyplot.imshow``.
:raises:
- ValueError
If y_axis is 'cqt_hz' or 'cqt_note' and fmin and fmax are not supplied.
'''
kwargs.setdefault('aspect', 'auto')
kwargs.setdefault('origin', 'lower')
kwargs.setdefault('interpolation', 'nearest')
if np.issubdtype(data.dtype, np.complex):
warnings.warn(
'Trying to display complex-valued input. Showing magnitude instead.'
)
data = np.abs(data)
kwargs.setdefault('cmap', cmap(data))
# NOTE: 2013-11-14 16:15:33 by Brian McFee <*****@*****.**>pitch
# We draw the image twice here. This is a hack to get around NonUniformImage
# not properly setting hooks for color: drawing twice enables things like
# colorbar() to work properly.
axes = plt.imshow(data, **kwargs)
if y_axis is 'log':
axes_phantom = plt.gca()
# Non-uniform imshow doesn't like aspect
del kwargs['aspect']
im_phantom = img.NonUniformImage(axes_phantom, **kwargs)
y_log, y_inv = __log_scale(data.shape[0])
im_phantom.set_data(np.arange(0, data.shape[1]), y_log, data)
axes_phantom.images.append(im_phantom)
axes_phantom.set_ylim(0, data.shape[0])
axes_phantom.set_xlim(0, data.shape[1])
# Set up the y ticks
positions = np.asarray(np.linspace(0, data.shape[0], n_yticks), dtype=int)
if y_axis is 'linear':
values = np.asarray(np.linspace(0, 0.5 * sr, data.shape[0] + 1),
dtype=int)
plt.yticks(positions, values[positions])
plt.ylabel('Hz')
elif y_axis is 'log':
values = np.asarray(np.linspace(0, 0.5 * sr, data.shape[0] + 1),
dtype=int)
plt.yticks(positions, values[y_inv[positions]])
plt.ylabel('Hz')
elif y_axis is 'mel':
m_args = {}
if fmin is not None:
m_args['fmin'] = fmin
if fmax is not None:
m_args['fmax'] = fmax
# only two star-args here, defined immediately above
values = librosa.core.mel_frequencies(
n_mels=data.shape[0], # pylint: disable=star-args
extra=True,
**m_args)[positions].astype(np.int)
plt.yticks(positions, values)
plt.ylabel('Hz')
elif y_axis is 'cqt_hz':
if fmax is None and fmin is None:
raise ValueError(
'fmin and fmax must be supplied for CQT axis display')
positions = np.arange(0,
data.shape[0],
np.ceil(data.shape[0] / float(n_yticks)),
dtype=int)
# Get frequencies
values = librosa.core.cqt_frequencies(
data.shape[0],
fmin=fmin,
bins_per_octave=int(data.shape[0] /
np.ceil(np.log2(fmax) - np.log2(fmin))))
plt.yticks(positions, values[positions].astype(int))
plt.ylabel('Hz')
elif y_axis is 'cqt_note':
if fmax is None and fmin is None:
raise ValueError(
'fmin and fmax must be supplied for CQT axis display')
positions = np.arange(0,
data.shape[0],
np.ceil(data.shape[0] / float(n_yticks)),
dtype=int)
# Get frequencies
values = librosa.core.cqt_frequencies(
data.shape[0],
fmin=fmin,
bins_per_octave=int(data.shape[0] /
np.ceil(np.log2(fmax) - np.log2(fmin))))
values = librosa.core.midi_to_note(
librosa.core.hz_to_midi(values[positions]))
plt.yticks(positions, values)
plt.ylabel('Note')
elif y_axis is 'chroma':
positions = np.arange(0, data.shape[0], max(1, data.shape[0] / 12))
# Labels start at 9 here because chroma starts at A.
values = librosa.core.midi_to_note(range(9, 9 + 12), octave=False)
plt.yticks(positions, values)
plt.ylabel('Pitch class')
elif y_axis is None or y_axis is 'off':
plt.yticks([])
plt.ylabel('')
else:
raise ValueError('Unknown y_axis parameter: %s' % y_axis)
# Set up the x ticks
positions = np.asarray(np.linspace(0, data.shape[1], n_xticks), dtype=int)
if x_axis is 'time':
time_ticks(positions,
librosa.core.frames_to_time(positions,
sr=sr,
hop_length=hop_length),
n_ticks=None,
axis='x')
plt.xlabel('Time')
elif x_axis is 'frames':
# Nothing to do here, plot is in frames
plt.xticks(positions, positions)
plt.xlabel('Frames')
elif x_axis is None or x_axis is 'off':
plt.xticks([])
plt.xlabel('')
else:
raise ValueError('Unknown x_axis parameter: %s' % x_axis)
return axes
x = scipy.ndimage.zoom(x, 3)
y = scipy.ndimage.zoom(y, 3)
data_fitted = scipy.ndimage.zoom(data_fitted.reshape(H.shape), 3)
plt.contour(x, y, data_fitted, 8, colors='w')
plt.ylim(yedges[0], yedges[-2])
plt.xlim(xedges[0], xedges[-2])
plt.savefig("measured_elec_pho_corr_%d.png" % (run_num))
plt.show()
raw_input()
plt.close()
fig2 = plt.figure()
ax2 = fig2.add_subplot(121, title='')
X, Y = np.meshgrid(xedges, yedges)
ax2.pcolormesh(X, Y, H)
ax2 = fig2.add_subplot(122, title='')
im2 = image.NonUniformImage(ax2, interpolation='bilinear')
xcenters = xedges[:-1] # + xedges[1:]) / 2
ycenters = yedges[:-1] # + yedges[1:]) / 2
im2.set_data(xcenters, ycenters, H)
ax2.images.append(im2)
# im2 = ax2.imshow(data_fitted, cmap=plt.cm.jet, origin='bottom',extent=(x.min(), x.max(),y.min(), y.max()))
ax2.contour(x, y, data_fitted, 8, colors='w')
plt.plot(
x_vals,
y_vals,
'w-',
label=
"rotation angle:%.1f$^{\circ}\pm$%.1f$^{\circ}$\npeak photon number:%d$\pm$%d\n peak electron number:%d$\pm$%d\n photon intercept:%d$\pm$%d\n electron intercept:%d$\pm$%d"
% (-np.rad2deg(np.arctan(slope)),
-np.rad2deg(np.arctan(slope) * rotation_err), max_pho,
max_pho * centerx_err, max_ele, max_ele * centery_err,