def _set_aspect_ratio(self, elem):
# hack to avoid bug with rasterize + invert_yaxis + aspect='equal'
# manually set data_aspect=1
options = elem.opts.get().options
frame_w = options.get('frame_width', None)
frame_h = options.get('frame_height', None)
if frame_w and frame_h:
# already set
return elem
wstart, wstop = elem.dimension_values(0, expanded=False)[[0, -1]]
hstart, hstop = elem.dimension_values(1, expanded=False)[[0, -1]]
w = wstop - wstart
h = hstop - hstart
if frame_w:
return elem.opts(
opts.Image(frame_height=int(round(frame_w / w * h))))
elif frame_h:
return elem.opts(
opts.Image(frame_width=int(round(frame_h / h * w))))
else:
return elem
def swatch(name, bounds=None, array=array, aspect=10, **kwargs):
"""Show swatch using matplotlib or bokeh via holoviews"""
if bounds is None:
bounds = (0, 0, 256, 1)
plot = hv.Image(array, bounds=bounds, group=name)
backends = hv.Store.loaded_backends()
if 'bokeh' in backends:
frame_height = kwargs.pop('frame_height', 60)
plot.opts(opts.Image(backend='bokeh', aspect=aspect, frame_height=frame_height,
toolbar='above', default_tools=['xwheel_zoom', 'xpan', 'save', 'reset'],
cmap=palette[name]))
if 'matplotlib' in backends:
def hook(plot, element):
plot.handles['axis'].axis('off')
plot.handles['axis'].set_title("sample", loc='left',
fontfamily='DejaVu Sans', fontsize=14,
fontweight='roman',fontstretch='semi-expanded')
fig_size = kwargs.pop('fig_size', 300)
plot.opts(opts.Image(backend='matplotlib', aspect=aspect, fig_size=fig_size,
title="", cmap=cm[name], hooks=[hook]))
return plot.opts(opts.Image(xaxis=None, yaxis=None), opts.Image(**kwargs))
def swatch(name, cmap=None, bounds=None, array=array, **kwargs):
"""Show swatch using matplotlib or bokeh via holoviews"""
title = name if cmap else get_aliases(name)
if bounds is None:
bounds = (0, 0, 256, 1)
cmap = list(cmap) if cmap is not None else None
plot = hv.Image(array, bounds=bounds, group=title)
backends = hv.Store.loaded_backends()
if 'bokeh' in backends:
width = kwargs.pop('width', 900)
height = kwargs.pop('height', 100)
plot.opts(
opts.Image(backend='bokeh',
width=width,
height=height,
toolbar='above',
default_tools=['xwheel_zoom', 'xpan', 'save', 'reset'],
cmap=cmap or palette[name]))
if 'matplotlib' in backends:
aspect = kwargs.pop('aspect', 15)
fig_size = kwargs.pop('fig_size', 350)
plot.opts(
opts.Image(backend='matplotlib',
aspect=aspect,
fig_size=fig_size,
cmap=cmap or cm[name]))
return plot.opts(opts.Image(xaxis=None, yaxis=None), opts.Image(**kwargs))
def test_plot_options_object_list(self):
im = Image(np.random.rand(10,10))
imopts1 = opts.Image(interpolation='nearest')
imopts2 = opts.Image(cmap='summer')
styled_im = im.options([imopts1,imopts2])
self.assertEqual(self.lookup_options(im, 'plot').options, {})
self.assertEqual(self.lookup_options(styled_im, 'style').options,
dict(cmap='summer', interpolation='nearest'))
def sine_comb(name, cmap=None, **kwargs):
"""Show sine_comb using matplotlib or bokeh via holoviews"""
title = name if cmap else get_aliases(name)
plot = hv.Image(sine, group=title)
backends = hv.Store.loaded_backends()
if 'bokeh' in backends:
plot.opts(opts.Image(backend='bokeh', width=400, height=150, toolbar='above',
cmap=cmap or palette[name]))
if 'matplotlib' in backends:
plot.opts(opts.Image(backend='matplotlib', aspect=3, fig_size=200,
cmap=cmap or cm[name]))
return plot.opts(opts.Image(xaxis=None, yaxis=None), opts.Image(**kwargs))
def panel(self, dmaps):
xy, zy, xz = dmaps
self._update_dynamic_values(xy, zy, xz)
zy.opts(
opts.Image(frame_width=self.frame_z_size,
frame_height=self.frame_y_size),
opts.RGB(frame_width=self.frame_z_size,
frame_height=self.frame_y_size),
)
if self.add_crosshairs:
self.get_crosshair()
panel_xy = self.z_viewer.panel(
(xy * self.xy_h * self.xy_v).relabel(group='orthoview'))
panel_zy = self.x_viewer.panel(
(zy * self.zy_h * self.zy_v).relabel(group='orthoview'))
panel_xz = self.y_viewer.panel(
(xz * self.xz_h * self.xz_v).relabel(group='orthoview'))
else:
panel_xy = self.z_viewer.panel(xy.relabel(group='orthoview'))
panel_zy = self.x_viewer.panel(zy.relabel(group='orthoview'))
panel_xz = self.y_viewer.panel(xz.relabel(group='orthoview'))
self._link_crosshairs()
return pn.Column(pn.Row(panel_xy, panel_zy),
pn.Row(panel_xz, self.param.navigaton_on))
def grid_plot(
self,
step: int,
cmap: Optional[Tuple] = None,
opts_list: Optional[opts.Image] = None,
):
"""
Plot the incoming data.
Graphical 2D-representation of the given (heigt x width) array.
Each site represents an individual speaking either language A,
language B, or language AB (bilinguals).
Args:
step: Integer value representing the iteration step during
the grid evolution.
cmap: N-Dimensional tuple containing the color palette used
to represent the different languages.
opts_list: Optional argument. opts.Image list specifying a
custom option list.
Returns:
The functions returns a holoviews image describing the
lattice, where each node is represented by the language
it speaks.
"""
data = self.memory[step]
opts_list = opts_list if opts_list else opts.Image(**self.return_plot_options(cmap))
return super(MinettWang, self).grid_plot(data=data, plot_option_list=opts_list)
def test_plot_options_one_object(self):
im = Image(np.random.rand(10,10))
imopts = opts.Image(interpolation='nearest', cmap='jet')
styled_im = im.options(imopts)
self.assertEqual(self.lookup_options(im, 'plot').options, {})
self.assertEqual(self.lookup_options(styled_im, 'style').options,
dict(cmap='jet', interpolation='nearest'))
def __post_init__(self):
"""
:return:
"""
data = self.spectral_cube.data
self.ds = hv.Dataset((np.arange(data.shape[2]), np.arange(
data.shape[1]), np.arange(data.shape[0]), data),
[self.spectral_axis_name, 'x', 'y'], 'Cube')
# maybe PolyEdit as well
# polys = hv.Polygons([hv.Box(int(self.image_width / 2), int(self.image_height / 2), int(self.image_height / 2))])
# self.box_stream = streams.PolyEdit(source=polys)
polys = hv.Polygons([])
self.box_stream = streams.BoxEdit(source=polys)
hlines = hv.HoloMap({i: hv.VLine(i)
for i in range(data.shape[2])}, 'wavelengths')
dmap = hv.DynamicMap(self.roi_curves, streams=[self.box_stream])
im = self.ds.to(hv.Image, ['x', 'y'], dynamic=True)
self.layout = (im * polys + dmap * hlines).opts(
opts.Image(cmap=self.color_map,
width=self.image_width,
height=self.image_height),
opts.Curve(width=650, height=450, framewise=True),
opts.Polygons(fill_alpha=0.2, line_color='white'),
opts.VLine(color='black'))
def test_opts_method_with_utility(self):
im = Image(np.random.rand(10,10))
imopts = opts.Image(cmap='Blues')
styled_im = im.opts(imopts)
assert styled_im is im
self.assertEqual(self.lookup_options(im, 'style').options,
{'cmap': 'Blues', 'interpolation': 'nearest'})
def swatch(name, cmap=None, bounds=None, array=array, **kwargs):
"""Show a color swatch for a colormap using matplotlib or bokeh via holoviews.
Colormaps can be selected by `name`, including those in Colorcet
along with any standard Bokeh palette or named Matplotlib colormap.
Custom colormaps can be visualized by passing an explicit
list of colors (for Bokeh) or the colormap object (for Matplotlib) to `cmap`.
HoloViews options for either backend can be passed in as kwargs,
so that you can customize the width, height, etc. of the swatch.
The `bounds` and `array` arguments allow you to customize the
portion of the colormap to show and how many samples to take
from it; see the source code and hv.Image documentation for
details.
"""
title = name if cmap else get_aliases(name)
if bounds is None:
bounds = (0, 0, 256, 1)
if type(cmap) is tuple:
cmap = list(cmap)
plot = hv.Image(array, bounds=bounds, group=title)
backends = hv.Store.loaded_backends()
if 'bokeh' in backends:
width = kwargs.pop('width', 900)
height = kwargs.pop('height', 100)
plot.opts(
opts.Image(backend='bokeh',
width=width,
height=height,
toolbar='above',
default_tools=['xwheel_zoom', 'xpan', 'save', 'reset'],
cmap=cmap or palette[name]))
if 'matplotlib' in backends:
aspect = kwargs.pop('aspect', 15)
fig_size = kwargs.pop('fig_size', 350)
plot.opts(
opts.Image(backend='matplotlib',
aspect=aspect,
fig_size=fig_size,
cmap=cmap or cm[name]))
return plot.opts(opts.Image(xaxis=None, yaxis=None), opts.Image(**kwargs))
def _plot_overlay(elevation, shadows):
'''
'''
shadows = hv.Dataset(
(np.arange(shadows.shape[2]), np.arange(
shadows.shape[0]), np.arange(shadows.shape[1]), shadows),
['Time', 'x', 'y'], 'Shadows')
elevation = hv.Dataset((np.arange(
elevation.shape[0]), np.arange(elevation.shape[1]), elevation),
['x', 'y'], 'Elevation')
opts.defaults(
opts.Image('elevation', cmap='viridis', invert_yaxis=True),
opts.Image('shadows', cmap='binary', invert_yaxis=True, alpha=0.7),
opts.Overlay(show_legend=False))
elevation = elevation.to(hv.Image, ['x', 'y'], group='elevation')
shadows = shadows.to(hv.Image, ['x', 'y'], group='shadows')
return elevation * shadows
def __init__(self, adh_mod, **params):
super(InterpolateMesh, self).__init__(adh_mod=adh_mod, **params)
# set defaults for initialized example
self.display_range.param.color_range.bounds = (10, 90)
self.display_range.color_range = (10, 90)
self.cmap_opts.colormap = cc.rainbow
self.scatter_projection.set_crs(ccrs.GOOGLE_MERCATOR)
self.adh_mod.wmts.source = gv.tile_sources.EsriImagery
# print(self.projection.param.UTM_zone_hemi.constant, self.projection.crs_label)
self.opts = (opts.Curve(height=self.map_height,
width=self.map_width,
xaxis=None,
line_width=1.50,
color='red',
tools=['hover']),
opts.Path(height=self.map_height,
width=self.map_width,
line_width=3,
color='black'),
opts.Image(height=self.map_height,
width=self.map_width,
cmap=self.cmap_opts.param.colormap,
clim=self.display_range.param.color_range,
colorbar=True,
clipping_colors={
'NaN': 'transparent',
'min': 'transparent'
},
axiswise=True),
opts.RGB(height=self.map_height, width=self.map_width),
opts.Points(height=self.map_height,
width=self.map_width,
color_index='z',
cmap=self.cmap_opts.param.colormap,
clim=self.display_range.param.color_range,
size=10,
tools=['hover'],
padding=(0.1, 0.1),
colorbar=True),
opts.TriMesh(height=self.map_height,
width=self.map_width,
color_index='z',
cmap=self.cmap_opts.param.colormap,
clim=self.display_range.param.color_range,
tools=['hover'],
padding=(0.1, 0.1),
colorbar=True), opts.VLine(color='black'))
def view(self):
data = self.data
xmin, xmax = np.min(data), np.max(data)
x1_dim = hv.Dimension('x₁', range=(xmin, xmax))
x2_dim = hv.Dimension('x₂', range=(xmin, xmax))
samples_map = self.maps["samples"]
samples_map = samples_map.opts(width=600,
height=350,
show_grid=True,
padding=(0, 0.1),
toolbar=None)
samples_map = samples_map.opts(opts.Path(color=blue, framewise=True))
samples_map = samples_map.redim.label(y="f(x)", x="x")
control_vline_map = self.maps["vlines_control"]
control_vline_map = control_vline_map.redim(x=x1_dim, y=x2_dim)
control_vline_map = control_vline_map.opts(show_grid=True)
control_vline_map = control_vline_map.opts(
opts.HLine(line_width=2), opts.VLine(line_width=2),
opts.Points(color="white", marker="s", size=8),
opts.Image(cmap="viridis"))
vlines_map = self.maps["vlines"]
vlines_map = vlines_map.opts(toolbar=None)
vlines_map = vlines_map.opts(opts.VLine(line_width=2),
opts.Points(size=6))
scatter_map = self.maps["scatter"]
scatter_map = scatter_map.redim.label(y="f(x₂)", x="f(x₁)")
scatter_map = scatter_map.opts(padding=(0.5, 0.5),
show_grid=True,
toolbar=None)
scatter_map = scatter_map.opts(
opts.Scatter(size=7,
framewise=True,
fill_color=orange1,
line_color=orange2))
title = pn.pane.Markdown("## GP samples visualization", max_height=25)
descr = pn.pane.Markdown(
"_For moving x₁ and x₂ bars: 1. turn off **pan** tool, 2. click and move the orange squared point on the covariance matrix_"
)
row0 = pn.Row(pn.Spacer(width=25), self.kernels_controller.view())
row1 = samples_map * vlines_map
row2 = pn.Row(control_vline_map, scatter_map)
return pn.Column(title, descr, row0, row1, row2)
def __post_init__(self):
"""
:return:
"""
self._pca_transform()
(shape_x, shape_y, n_component) = self.scores_cube.shape
self.ds = hv.Dataset((np.arange(n_component), np.arange(shape_y),
np.arange(shape_x), self.scores_cube),
['Principal Component', 'x', 'y'], 'Cube')
im = self.ds.to(hv.Image, ['x', 'y'], dynamic=True)
self.layout = im.opts(
opts.Image(cmap=self.color_map,
width=self.image_width,
height=self.image_height))
def view_map(self):
# print('view_map method')
if self.adh_mod.mesh.elevation_toggle:
elevation = rasterize(self.adh_mod.mesh.tri_mesh,
aggregator=ds.mean('z'),
precompute=True).apply.opts(
opts.Image(
cmap=self.cmap_opts.colormap,
clim=self.display_range.color_range,
height=self.map_height,
width=self.map_width))
else:
elevation = Curve([]).opts(height=self.map_height,
width=self.map_width)
# return self.adh_mod.mesh.view_bathy() * self.adh_mod.mesh.view_elements(line_color='yellow') * base_map * self.view_scatter()
return elevation * self.adh_mod.mesh.view_elements(
line_color='yellow') * hv.DynamicMap(
self.adh_mod.wmts.view) * self.view_scatter()
# In[ ]
import holoviews as hv, datashader as ds
from holoviews import opts
import dask.dataframe as dd
from holoviews.operation.datashader import rasterize
hv.extension('bokeh')
# In[ ]
path = './nyc_taxi_wide.parq'
df = dd.read_parquet(path).persist()
points = hv.Points(df, ['dropoff_x', 'dropoff_y'],
['dropoff_hour', 'pickup_hour'])
# In[ ]
options = opts.Image(data_aspect=1,
responsive=True,
logz=True,
tools=['hover'],
colorbar=True)
rasterize(points).options(options)
# In[ ]
import panel
panel.panel(rasterize(points).options(options))
# -*- coding: utf-8 -*-
"""Tools to work with images.
"""
import circle_fit as cf
import holoviews as hv
import numpy as np
import pandas as pd
from astropy.io import fits
from holoviews import opts
from skimage.exposure import equalize_adapthist as equalize
from skimage.exposure import rescale_intensity
hv.extension("bokeh")
opts.defaults(
opts.Image(tools=["hover"], cmap="gray"),
opts.Points(color="red", marker="x", size=20),
)
class Image:
def __init__(self, path):
self.path = path
self.data, self.header = fits.getdata(path, header=True)
@property
def name(self):
return self.path.name
@property
def wavelength(self):
return self.header["FILTER"]
import panel as pn
from skimage.exposure import rescale_intensity
from holoviews import opts
from holoviews.operation.datashader import rasterize
from inter_view.color import available_cmaps
from inter_view.utils import blend_overlay
# defines default options for all viewers
opts.defaults(
opts.Image('channel',
frame_width=600,
invert_yaxis=True,
xaxis='bare',
yaxis='bare',
bgcolor='black',
active_tools=['pan', 'wheel_zoom'],
show_title=False),
opts.RGB('composite',
frame_width=600,
invert_yaxis=True,
xaxis='bare',
yaxis='bare',
bgcolor='black',
active_tools=['pan', 'wheel_zoom'],
show_title=False),
opts.HLine('orthoview',
line_dash='dashed',
line_width=1,
line_color='white'),
def _plot(self) -> pn.panel:
"""
Represent the initial and final state of the lattice.
Graphical representation of the lattice. The image shows the
initial and final state of the grid (in order to compare how
the network has evolved), as well as the number of speakers
as a function of time. self.track = True is needed to call this
method.
"""
grid_flat = self.memory.reshape(self.memory.shape[0], -1)
speakers_a = (grid_flat == 1).sum(1)
speakers_b = (grid_flat == -1).sum(1)
speakers_ab = (grid_flat == 0).sum(1)
total = speakers_a + speakers_b + speakers_ab == (self.width * self.height) * np.ones(
len(self.memory),
)
if not np.all(total):
raise ValueError(
"The total number of speakers does not correspond to the lattice size!",
)
# Plots
colors = ["navy", "white", "red"]
data_start = self.memory[0]
data_end = self.grid.data
grid_start = {
"xdata": np.arange(1, data_start.shape[1] + 1),
"ydata": np.arange(1, data_start.shape[0] + 1),
"zdata": data_start,
}
grid_end = {
"xdata": np.arange(1, data_end.shape[1] + 1),
"ydata": np.arange(1, data_end.shape[0] + 1),
"zdata": data_end,
}
plot_start = hv.Image(
grid_start,
kdims=["xdata", "ydata"],
vdims=hv.Dimension("zdata", range=(-1, 1)),
label="Initial grid",
)
plot_end = hv.Image(
grid_end,
kdims=["xdata", "ydata"],
vdims=hv.Dimension("zdata", range=(-1, 1)),
label="Final grid",
)
plot_curvea = hv.Curve(speakers_a, label="Speakers A").opts(color="red")
plot_curveb = hv.Curve(speakers_b, label="Speakers B").opts(color="navy")
plot_curveab = hv.Curve(speakers_ab, label="Speakers AB").opts(color="gray")
# Compositions
grids = plot_start + plot_end
lines = plot_curvea * plot_curveb * plot_curveab
layout = grids + lines
# Options
layout.opts(
opts.Image(
invert_yaxis=True,
cmap=colors,
colorbar=True,
width=350,
labelled=[],
colorbar_opts={
"title": "Languages",
"title_text_align": "left",
"major_label_overrides": {-1: "B", 0: "AB", 1: "A"},
"ticker": FixedTicker(ticks=[-1, 0, 1]),
"major_label_text_align": "right",
},
),
opts.Curve(xlabel="Iterations", ylabel="Number of speakers", width=700),
)
return display(pn.Column(pn.Row(plot_start, plot_end), lines))
def plot_cubes(cube, mode='slider', backend='matplotlib', dpi=100,
figtype='png', vmin=None, vmax=None, size=100, width=360,
height=360, cmap=None, colorbar=True, dynamic=True,
anim_path=None, data_step_range=None, label=None,
label_step_range=None, delay=50, anim_format='gif',
delete_anim_cache=True, **kwargs):
""" Plot multi-dimensional high-contrast imaging datacubes (3d and 4d
``numpy`` arrays). It allows to visualize in-memory ``numpy`` arrays on
``Jupyterlab`` by leveraging the ``HoloViews`` library. It can also generate
and save animations from a 3d ``numpy`` array with ``matplotlib``.
Parameters
----------
cube : np.ndarray
Input 3d or 4d cube.
mode : {'slider', 'animation'}, str optional
Whether to plot the 3d array as a widget with a slider or to save an
animation of the 3d array. The animation is saved to disk using
ImageMagick's convert command (it must be installed otherwise a
``FileNotFoundError`` will be raised).
backend : {'matplotlib', 'bokeh'}, str optional
Selects the backend used to display the plots. ``Bokeh`` plots are
interactive, allowing the used to zoom, pan, inspect pixel values, etc.
``Matplotlib`` can lead to some flickering when using the slider and
``dynamic`` is True.
dpi : int, optional
[backend='matplotlib'] The rendered dpi of the figure.
figtype : {'png', 'svg'}, str optional
[backend='matplotlib'] Type of output.
vmin : None, float or int, optional
For defining the data range that the colormap covers. When set to None,
the colormap covers the complete value range of the supplied data.
vmax : None, float or int, optional
For defining the data range that the colormap covers. When set to None,
the colormap covers the complete value range of the supplied data.
size : int, optional
[backend='matplotlib'] Sets the size of the plot.
width : int, optional
[backend='bokeh'] Sets the width of the plot.
height : int, optional
[backend='bokeh'] Sets the height of the plot.
cmap : None or str, optional
Colormap. When None, the value of the global variable ``default_cmap``
will be used.
colorbar : bool, optional
If True, a colorbar is shown.
dynamic : bool, optional
[mode='slider'] When False, a ``HoloViews.HoloMap`` is created (slower
and will take up a lot of RAM for large datasets). If True, a
``HoloViews.DynamicMap`` is created instead.
anim_path : str, optional
[mode='animation'] The animation path/filename. If None then the
animation will be called ``animation``.``anim_format`` and will be saved
in the current directory.
data_step_range : tuple, optional
[mode='animation'] Tuple of 1, 2 or 3 values that creates a range for
slicing the ``data`` cube.
label : str, optional
[mode='animation'] Label to be overlaid on top of each frame of the
animation. If None, then ``frame <#>`` will be used.
label_step_range : tuple, optional
[mode='animation'] Tuple of 1, 2 or 3 values that creates a range for
customizing the label overlaid on top of each frame of the animation.
delay : int, optional
[mode='animation'] Delay for displaying the frames in the animation
sequence.
anim_format : str, optional
[mode='animation'] Format of the saved animation. By default 'gif' is
used. Other formats supported by ImageMagick are valid, such as 'mp4'.
delete_anim_cache : str, optional
[mode='animation'] If True, the cache folder is deleted once the
animation file is saved to disk.
**kwargs : dictionary, optional
[mode='animation'] Arguments to be passed to ``plot_frames`` to
customize each frame of the animation (adding markers, using a log
scale, etc).
Notes
-----
http://holoviews.org/getting_started/Gridded_Datasets.html
http://holoviews.org/user_guide/Gridded_Datasets.html
http://holoviews.org/user_guide/Applying_Customizations.html
"""
hv.extension(backend)
if not isinstance(cube, np.ndarray):
raise TypeError('`cube` must be a numpy.ndarray')
if cmap is None:
cmap = default_cmap
if mode == 'slider':
if cube.ndim not in (3, 4):
raise ValueError('`cube` must be a 3 or 4 array when `mode` set to '
'slider')
if cube.ndim == 3:
# Dataset((X, Y, Z), Data), where
# X is a 1D array of shape M ,
# Y is a 1D array of shape N and
# Z is a 1D array of shape O
# Data is a ND array of shape NxMxO
ds = hv.Dataset((range(cube.shape[2]), range(cube.shape[1]),
range(cube.shape[0]), cube), ['x', 'y', 'time'],
'flux')
max_frames = cube.shape[0]
elif cube.ndim == 4:
# adding a lambda dimension
ds = hv.Dataset((range(cube.shape[3]), range(cube.shape[2]),
range(cube.shape[1]), range(cube.shape[0]), cube),
['x', 'y', 'time', 'lambda'], 'flux')
max_frames = cube.shape[0] * cube.shape[1]
# Matplotlib takes None but not Bokeh. We take global min & max instead
if vmin is None:
vmin = cube.min()
if vmax is None:
vmax = cube.max()
print(ds)
print(":Cube_shape\t{}".format(list(cube.shape[::-1])))
# not working for bokeh: dpi
image_stack = ds.to(hv.Image, kdims=['x', 'y'], dynamic=dynamic)
hv.output(backend=backend, size=size, dpi=dpi, fig=figtype,
max_frames=max_frames)
if backend == 'matplotlib':
# keywords in the currently active 'matplotlib' renderer are:
# 'alpha', 'clims', 'cmap', 'filterrad', 'interpolation', 'norm',
# 'visible'
#options = "Image (cmap='" + cmap + "', interpolation='nearest',"
#options += " clims=("+str(vmin)+','+str(vmax)+")"+")"
#opts(options, image_stack)
return image_stack.opts(opts.Image(colorbar=colorbar,
cmap=cmap,
clim=(vmin, vmax)))
# hv.save(image_stack, 'holomap.gif', fps=5)
elif backend == 'bokeh':
#options = "Image (cmap='" + cmap + "')"
#opts(options, image_stack)
# Compensating the width to accommodate the colorbar
if colorbar:
cb_wid = 15
cb_pad = 3
tick_len = len(str(int(cube.max())))
if tick_len < 4:
cb_tick = 25
elif tick_len == 4:
cb_tick = 35
elif tick_len > 4:
cb_tick = 45
width_ = width + cb_pad + cb_wid + cb_tick
else:
width_ = width
return image_stack.opts(opts.Image(colorbar=colorbar,
colorbar_opts={'width': 15,
'padding': 3},
width=width_, height=height,
clim=(vmin, vmax),
cmap=cmap,
tools=['hover']))
elif mode == 'animation':
if cube.ndim != 3:
raise ValueError('`cube` must be a 3 array when `mode` set to '
'animation')
if backend == 'bokeh':
print('Creating animations works with the matplotlib backend')
dir_path = './animation_temp/'
if anim_path is None:
anim_path = './animation'
if data_step_range is None:
data_step_range = range(0, cube.shape[0], 1)
else:
if not isinstance(data_step_range, tuple):
msg = '`data_step_range` must be a tuple with 1, 2 or 3 values'
raise ValueError(msg)
if len(data_step_range) == 1:
data_step_range = range(data_step_range)
elif len(data_step_range) == 2:
data_step_range = range(data_step_range[0], data_step_range[1])
elif len(data_step_range) == 3:
data_step_range = range(data_step_range[0],
data_step_range[1],
data_step_range[2])
if label_step_range is None:
label_step_range = data_step_range
else:
if not isinstance(label_step_range, tuple):
msg = '`label_step_range` must be a tuple with 1, 2 or 3 values'
raise ValueError(msg)
if len(label_step_range) == 1:
label_step_range = range(label_step_range)
elif len(label_step_range) == 2:
label_step_range = range(label_step_range[0],
label_step_range[1])
elif len(label_step_range) == 3:
label_step_range = range(label_step_range[0],
label_step_range[1],
label_step_range[2])
if os.path.exists(dir_path):
shutil.rmtree(dir_path)
print('Replacing ' + dir_path)
os.mkdir(dir_path)
print('Producing each animation frame...')
for i, labstep in zip(data_step_range, list(label_step_range)):
if label is None:
label = 'frame '
savelabel = dir_path + label + str(i + 100)
plot_frames(cube[i], backend='matplotlib', mode='mosaic',
save=savelabel, dpi=dpi, vmin=vmin, vmax=vmax,
colorbar=colorbar, cmap=cmap,
label=[label + str(labstep + 1)], **kwargs)
try:
filename = anim_path + '.' + anim_format
call(['convert', '-delay', str(delay), dir_path + '*.png',
filename])
if os.path.exists(filename):
print('Animation successfully saved to disk as ' + filename)
if delete_anim_cache:
shutil.rmtree(dir_path)
print('Temp directory deleted ' + dir_path)
except FileNotFoundError:
print('ImageMagick `convert` command could not be found')
else:
raise ValueError("`mode` is not recognized")
def get_fractal(x_range, y_range):
(x0, x1), (y0, y1) = x_range, y_range
image = np.zeros((600, 600), dtype=np.uint8)
return hv.Image(create_fractal(x0, x1, -y1, -y0, image, 200),
bounds=(x0, y0, x1, y1))
# Define stream linked to axis XY-range
range_stream = RangeXY(x_range=(-1., 1.), y_range=(-1., 1.))
# Create DynamicMap to compute fractal per zoom range and
# adjoin a logarithmic histogram
dmap = hv.DynamicMap(get_fractal,
label='Manderbrot Explorer',
streams=[range_stream]).hist(log=True)
# Apply options
dmap.opts(
opts.Histogram(framewise=True, logy=True, width=200),
opts.Image(cmap='fire',
logz=True,
height=600,
width=600,
xaxis=None,
yaxis=None))
doc = renderer.server_doc(dmap)
doc.title = 'Mandelbrot Explorer'
def __call__(self, dset, **params):
self.p = ParamOverrides(self, params)
if self.p.vdim is None:
vdim = dset.vdims[0].name
else:
vdim = self.p.vdim
ra_range = (ra0, ra1) = dset.range("ra")
if self.p.ra_sampling:
xsampling = (ra1 - ra0) / self.p.ra_sampling
else:
xsampling = None
dec_range = (dec0, dec1) = dset.range("dec")
if self.p.dec_sampling:
ysampling = (dec1 - dec0) / self.p.dec_sampling
else:
ysampling = None
if self.p.aggregator == "mean":
aggregator = ds.mean(vdim)
elif self.p.aggregator == "std":
aggregator = ds.std(vdim)
elif self.p.aggregator == "count":
aggregator = ds.count()
sky_range = RangeXY()
if self.p.range_stream:
def redim(dset, x_range, y_range):
ranges = {}
if x_range and all(isfinite(v) for v in x_range):
ranges["ra"] = x_range
if y_range and all(isfinite(v) for v in x_range):
ranges["dec"] = y_range
return dset.redim.range(**ranges) if ranges else dset
dset = dset.apply(redim, streams=[self.p.range_stream])
link_streams(self.p.range_stream, sky_range)
streams = [sky_range, PlotSize()]
pts = dset.apply(skypoints, streams=[self.p.filter_stream])
reset = PlotReset(source=pts)
reset.add_subscriber(partial(reset_stream, None,
[self.p.range_stream]))
rasterize_inst = rasterize.instance(aggregator=aggregator,
streams=streams,
x_sampling=xsampling,
y_sampling=ysampling)
raster_pts = apply_when(
pts,
operation=rasterize_inst,
predicate=lambda pts: len(pts) > self.p.max_points)
return raster_pts.opts(
opts.Image(
bgcolor="black",
colorbar=True,
cmap=self.p.cmap,
min_height=100,
responsive=True,
tools=["hover"],
symmetric=True,
),
opts.Points(
color=vdim,
cmap=self.p.cmap,
framewise=True,
size=self.p.decimate_size,
tools=["hover"],
symmetric=True,
),
opts.Overlay(hooks=[
partial(reset_hook, x_range=ra_range, y_range=dec_range)
]),
)
def __call__(self, dset, **params):
self.p = ParamOverrides(self, params)
if self.p.xdim not in dset.dimensions():
raise ValueError("{} not in Dataset.".format(self.p.xdim))
if self.p.ydim not in dset.dimensions():
raise ValueError("{} not in Dataset.".format(self.p.ydim))
if ("ra" not in dset.dimensions()) or ("dec" not in dset.dimensions()):
raise ValueError("ra and/or dec not in Dataset.")
# Compute sampling
ra_range = (ra0, ra1) = dset.range("ra")
if self.p.ra_sampling:
ra_sampling = (ra1 - ra0) / self.p.x_sampling
else:
ra_sampling = None
dec_range = (dec0, dec1) = dset.range("dec")
if self.p.dec_sampling:
dec_sampling = (dec1 - dec0) / self.p.y_sampling
else:
dec_sampling = None
x_range = (x0, x1) = dset.range(self.p.xdim)
if self.p.x_sampling:
x_sampling = (x1 - x0) / self.p.x_sampling
else:
x_sampling = None
y_range = (y0, y1) = dset.range(self.p.ydim)
if self.p.y_sampling:
y_sampling = (y1 - y0) / self.p.y_sampling
else:
y_sampling = None
# Set up scatter plot
scatter_range = RangeXY()
if self.p.scatter_range_stream:
def redim_scatter(dset, x_range, y_range):
ranges = {}
if x_range and all(isfinite(v) for v in x_range):
ranges[self.p.xdim] = x_range
if y_range and all(isfinite(v) for v in x_range):
ranges[self.p.ydim] = y_range
return dset.redim.range(**ranges) if ranges else dset
dset_scatter = dset.apply(redim_scatter,
streams=[self.p.scatter_range_stream])
link_streams(self.p.scatter_range_stream, scatter_range)
else:
dset_scatter = dset
scatter_pts = dset_scatter.apply(filterpoints,
streams=[self.p.filter_stream],
xdim=self.p.xdim,
ydim=self.p.ydim)
scatter_streams = [scatter_range, PlotSize()]
scatter_rasterize = rasterize.instance(streams=scatter_streams,
x_sampling=x_sampling,
y_sampling=y_sampling)
cmap = (process_cmap(self.p.scatter_cmap)[:250]
if self.p.scatter_cmap == "fire" else self.p.scatter_cmap)
scatter_rasterized = apply_when(
scatter_pts,
operation=scatter_rasterize,
predicate=lambda pts: len(pts) > self.p.max_points
).opts(
opts.Image(clim=(1, np.nan),
clipping_colors={"min": "transparent"},
cmap=cmap),
opts.Points(clim=(1, np.nan),
clipping_colors={"min": "transparent"},
cmap=cmap),
opts.Overlay(
hooks=[partial(reset_hook, x_range=x_range, y_range=y_range)]),
)
# Set up sky plot
sky_range = RangeXY()
if self.p.sky_range_stream:
def redim_sky(dset, x_range, y_range):
ranges = {}
if x_range and all(isfinite(v) for v in x_range):
ranges["ra"] = x_range
if y_range and all(isfinite(v) for v in x_range):
ranges["dec"] = y_range
return dset.redim.range(**ranges) if ranges else dset
dset_sky = dset.apply(redim_sky, streams=[self.p.sky_range_stream])
link_streams(self.p.sky_range_stream, sky_range)
else:
dset_sky = dset
sky_pts = dset_sky.apply(filterpoints,
xdim="ra",
ydim="dec",
set_title=False,
streams=[self.p.filter_stream])
skyplot_streams = [sky_range, PlotSize()]
sky_rasterize = rasterize.instance(
aggregator=ds.mean(self.p.ydim),
streams=skyplot_streams,
x_sampling=ra_sampling,
y_sampling=dec_sampling,
)
sky_rasterized = apply_when(
sky_pts,
operation=sky_rasterize,
predicate=lambda pts: len(pts) > self.p.max_points).opts(
opts.Image(bgcolor="black",
cmap=self.p.sky_cmap,
symmetric=True),
opts.Points(bgcolor="black",
cmap=self.p.sky_cmap,
symmetric=True),
opts.Overlay(hooks=[
partial(reset_hook, x_range=ra_range, y_range=dec_range)
]),
)
# Set up BoundsXY streams to listen to box_select events and notify FilterStream
scatter_select = BoundsXY(source=scatter_pts)
scatter_notifier = partial(notify_stream,
filter_stream=self.p.filter_stream,
xdim=self.p.xdim,
ydim=self.p.ydim)
scatter_select.add_subscriber(scatter_notifier)
sky_select = BoundsXY(source=sky_pts)
sky_notifier = partial(notify_stream,
filter_stream=self.p.filter_stream,
xdim="ra",
ydim="dec")
sky_select.add_subscriber(sky_notifier)
# Reset
reset = PlotReset(source=sky_pts)
reset.add_subscriber(
partial(reset_stream, self.p.filter_stream,
[self.p.sky_range_stream, self.p.scatter_range_stream]))
raw_scatterpts = filterpoints(dset, xdim=self.p.xdim, ydim=self.p.ydim)
raw_scatter = datashade(
raw_scatterpts,
cmap=list(Greys9[::-1][:5]),
streams=scatter_streams,
x_sampling=x_sampling,
y_sampling=y_sampling,
)
scatter_p = raw_scatter * scatter_rasterized
if self.p.show_rawsky:
raw_skypts = filterpoints(dset, xdim=self.p.xdim, ydim=self.p.ydim)
raw_sky = datashade(
rawskypts,
cmap=list(Greys9[::-1][:5]),
streams=skyplot_streams,
x_sampling=ra_sampling,
y_sampling=dec_sampling,
)
sky_p = raw_sky * sky_rasterized
else:
sky_p = sky_rasterized
if self.p.show_table:
table = dset.apply(summary_table,
ydim=self.p.ydim,
streams=[self.p.filter_stream])
table = table.opts()
layout = table + scatter_p + sky_p
else:
layout = (scatter_p + sky_p).opts(sizing_mode="stretch_width")
return layout.opts(
opts.Image(colorbar=True,
responsive=True,
tools=["box_select", "hover"]),
opts.Layout(sizing_mode="stretch_width"),
opts.Points(color=self.p.ydim, tools=["hover"]),
opts.RGB(alpha=0.5),
opts.Table(width=200),
)
img.data = step(img.data)
return hv.Image(img)
# Set up plot which advances on counter and adds pattern on tap
title = 'Game of Life - Tap to place pattern, Doubletap to clear'
img = hv.Image(np.zeros((100, 200), dtype=np.uint8))
counter, tap = Counter(transient=True), Tap(transient=True),
pattern_dim = hv.Dimension('Pattern', values=sorted(shapes.keys()))
dmap = hv.DynamicMap(update, kdims=[pattern_dim], streams=[counter, tap])
plot = dmap.opts(
opts.Image(cmap='gray',
clim=(0, 1),
toolbar=None,
responsive=True,
min_height=800,
title=title,
xaxis=None,
yaxis=None))
# Add callback to clear on double tap
def reset_data(x, y):
img.data[:] = 0
reset = DoubleTap(transient=True, source=plot)
reset.add_subscriber(reset_data)
# Set up Panel app and periodic callback
panel = pn.pane.HoloViews(plot, center=True, widget_location='right')
config = yaml.load(fd.read())
# configure logging
logging.config.dictConfig(config["logging"])
# load bokeh
hv.extension("bokeh")
pn.extension(sizing_mode="scale_width")
# Set some defaults for the visualization of the graphs
hvopts.defaults(
hvopts.Image( # pylint: disable=no-member
# Don't set both height and width, or the UI will not be responsive!
height=600,
# width=800,
responsive=True,
show_title=True,
tools=["hover"],
active_tools=["pan", "wheel_zoom"],
align="end",
),
hvopts.Layout(toolbar="right"), # pylint: disable=no-member
)
ui = thalassa.ThalassaUI(
display_variables=True,
display_stations=True,
)
# https://panel.holoviz.org/reference/templates/Bootstrap.html
bootstrap = pn.template.BootstrapTemplate(
site="example.com",
def view(self,
plane='axial',
three_planes=False,
image_size=300,
dynamic=True,
cmap='gray'):
# imopts = {'tools': ['hover'], 'width': 400, 'height': 400, 'cmap': 'gray'}
# imopts = {'tools': ['hover'], 'cmap': 'gray'}
opts.defaults(
opts.GridSpace(
shared_xaxis=False,
shared_yaxis=False,
fontsize={
'title': 16,
'labels': 16,
'xticks': 12,
'yticks': 12
},
),
# opts.Image(cmap='gray', tools=['hover'], xaxis=None,
# yaxis=None, shared_axes=False),
# opts.Overlay(tools=['hover']),
# opts.NdOverlay(tools=['hover']),
opts.Image(cmap=cmap, xaxis=None, yaxis=None, shared_axes=False),
)
self.is2d = False
if 'z' not in self.ds.dims:
self.is2d = True
self.set_size(image_size)
if self.is2d:
plane == '2d'
a1, a2 = 'x', 'y'
pane_width = self.pane_width
pane_height = self.pane_height
else:
if plane == 'axial':
a1, a2, a3 = 'x', 'y', 'z'
# invert = True
pane_width = self.axial_width
pane_height = self.axial_height
elif plane == 'coronal':
a1, a2, a3 = 'x', 'z', 'y'
pane_width = self.coronal_width
pane_height = self.coronal_height
# invert = False
elif plane == 'sagittal':
a1, a2, a3 = 'y', 'z', 'x'
# invert = False
pane_width = self.sagittal_width
pane_height = self.sagittal_height
contrast_start_min = np.asscalar(
self.ds.isel(subject_id=0, ).image.quantile(0.01).values) - 1e-6
contrast_start_max = np.asscalar(
self.ds.isel(subject_id=0, ).image.quantile(0.99).values) + 1e-6
contrast_min = np.asscalar(
self.ds.isel(subject_id=0).image.min().values)
contrast_max = np.asscalar(
self.ds.isel(subject_id=0).image.max().values)
ctotal = contrast_max - contrast_min
contrast_min -= ctotal * 0.1
contrast_max += ctotal * 0.1
cslider = pn.widgets.RangeSlider(start=contrast_min,
end=contrast_max,
value=(contrast_start_min,
contrast_start_max),
name='contrast')
if 'overlay' in self.ds.data_vars:
hv_ds_image = hv.Dataset(self.ds[['image', 'overlay']])
# hv_ds_image = hv.Dataset(self.ds['image'])
if self.verbose:
print(hv_ds_image)
hv_ds_overlay = hv.Dataset(self.ds['overlay'])
if self.verbose:
print(hv_ds_overlay)
# tooltips = [
# ('(x,y)', '($x, $y)'),
# ('image', '@image'),
# ('overlay', '@overlay')
# ]
# hover = HoverTool(tooltips=tooltips)
if self.verbose:
print('overlay_max_calc')
if self.is2d:
first_subj_max = self.ds.isel(subject_id=0).overlay.max(
dim=['x', 'y', 'label']).compute()
first_subj_min = self.ds.isel(subject_id=0).overlay.min(
dim=['x', 'y', 'label']).compute()
else:
first_subj_max = self.ds.isel(subject_id=0).overlay.max(
dim=['x', 'y', 'z', 'label']).compute()
first_subj_min = self.ds.isel(subject_id=0).overlay.min(
dim=['x', 'y', 'z', 'label']).compute()
if self.verbose:
print('overlay_max_calc ready')
print(first_subj_max)
overlay_max = first_subj_max.max()
alpha_slider = pn.widgets.FloatSlider(start=0,
end=1,
value=0.7,
name='overlay transparency')
cmap_select = pn.widgets.Select(name='Overlay Colormap',
options=['Discrete', 'Continuous'])
if self.verbose:
print('max thresh calc')
print(first_subj_max.max())
max_thresholds = first_subj_max.values
if max_thresholds.size != 1:
max_thresholds = sorted(set(max_thresholds))
else:
max_thresholds = [np.asscalar(max_thresholds)]
# max_thresholds = sorted(list(set([first_subj.overlay.sel(overlay_label=i).values.max()
# for i in first_subj.overlay_label])))
if self.verbose:
print('min thresh calc')
min_thresholds = first_subj_min.values + 1e-6
if min_thresholds.size != 1:
min_thresholds = sorted(set(min_thresholds))
else:
min_thresholds = [np.asscalar(min_thresholds)]
# min_thresholds = sorted(list(set(first_subj_min.min())))
# min_thresholds = sorted(list(set([first_subj.sel(overlay_label=i).min()+1e-6 for i in
# first_subj.overlay_label])))
# ocslider = pn.widgets.DiscreteSlider(name='overlay max threshold',
# options=max_thresholds,
# value=max_thresholds[-1])
if len(min_thresholds) == 1 and len(max_thresholds) == 1:
thresh_toggle = 0
oclim = (min_thresholds[0], max_thresholds[0])
elif len(min_thresholds) > 1 and len(max_thresholds) == 1:
thresh_toggle = 1
ocslider_min = pn.widgets.DiscreteSlider(
name='overlay min threshold',
options=min_thresholds,
value=min_thresholds[-1])
@pn.depends(ocslider_min)
def oclim(value):
return (value, max_thresholds[0])
elif len(min_thresholds) == 1 and len(max_thresholds) > 1:
thresh_toggle = 2
ocslider_max = pn.widgets.DiscreteSlider(
name='overlay max threshold',
options=max_thresholds,
value=max_thresholds[-1])
@pn.depends(ocslider_max)
def oclim(value):
return (min_thresholds[0], value)
else:
thresh_toggle = 3
ocslider_min = pn.widgets.DiscreteSlider(
name='overlay min threshold',
options=min_thresholds,
value=min_thresholds[-1])
ocslider_max = pn.widgets.DiscreteSlider(
name='overlay max threshold',
options=max_thresholds,
value=max_thresholds[-1])
@pn.depends(ocslider_min, ocslider_max)
def oclim(value_min, value_max):
return (value_min, value_max)
if self.verbose:
print(thresh_toggle)
@pn.depends(cmap_select)
def cmap_dict(value):
d = {'Discrete': 'glasbey_hv', 'Continuous': 'viridis'}
return d[value]
# subj_viewer = SubjectViewer(ds=self.ds,
# subject_id_sel=list(self.ds.subject_id.values))
if self.is2d:
gridspace = hv_ds_image.to(
hv.Image, [a1, a2],
vdims=['image', 'overlay'],
dynamic=dynamic).opts(
frame_width=pane_width,
frame_height=pane_height,
tools=['hover'],
).apply.opts(clim=cslider.param.value)
if self.verbose:
print(gridspace)
gridspace *= hv_ds_overlay.to(
hv.Image, [a1, a2], vdims='overlay', dynamic=dynamic).opts(
cmap='glasbey_hv',
clipping_colors={
'min': 'transparent',
'NaN': 'transparent'
},
).redim.range(overlay=(1e-6, overlay_max)).apply.opts(
alpha=alpha_slider.param.value,
cmap=cmap_dict,
clim=oclim)
# print(gridspace)
# print(gridspace)
# gridspace = hv.DynamicMap(subj_viewer.load_subject).grid('label')
gridspace = gridspace.layout('label')
elif three_planes:
# squish_height = int(max(image_size*(len(self.ds.z)/len(self.ds.x)), image_size/2))
# gridspace = hv.GridSpace(kdims=['plane', 'label'], label=f'{self.subject_id}')
gridspace = hv.GridSpace(kdims=['plane', 'label'])
for mod in self.ds.label.values:
gridspace['axial', mod] = hv_ds_image.select(label=mod).to(
hv.Image, ['x', 'y'],
groupby=['subject_id', 'z'],
vdims='image',
dynamic=dynamic).opts(
frame_width=self.axial_width,
frame_height=self.axial_height).apply.opts(
clim=cslider.param.value)
gridspace['coronal', mod] = hv_ds_image.select(
label=mod).to(
hv.Image, ['x', 'z'],
groupby=['subject_id', 'y'],
vdims='image',
dynamic=dynamic).opts(
frame_width=self.coronal_width,
frame_height=self.coronal_height).apply.opts(
clim=cslider.param.value)
gridspace['sagittal', mod] = hv_ds_image.select(
label=mod).to(
hv.Image, ['y', 'z'],
groupby=['subject_id', 'x'],
vdims='image',
dynamic=dynamic).opts(
frame_width=self.sagittal_width,
frame_height=self.sagittal_height).apply.opts(
clim=cslider.param.value)
gridspace['axial',
mod] *= hv_ds_overlay.select(label=mod).to(
hv.Image, ['x', 'y'],
groupby=['subject_id', 'z', 'overlay_label'],
vdims='overlay',
dynamic=dynamic).opts(
cmap='glasbey_hv',
clipping_colors={
'min': 'transparent',
'NaN': 'transparent'
},
).redim.range(
overlay=(0.1, overlay_max)).apply.opts(
alpha=alpha_slider.param.value,
cmap=cmap_dict,
clim=oclim)
gridspace['coronal',
mod] *= hv_ds_overlay.select(label=mod).to(
hv.Image, ['x', 'z'],
groupby=['subject_id', 'y', 'overlay_label'],
vdims='overlay',
dynamic=dynamic).opts(
cmap='glasbey_hv',
clipping_colors={
'min': 'transparent',
'NaN': 'transparent'
},
).redim.range(
overlay=(0.1, overlay_max)).apply.opts(
alpha=alpha_slider.param.value,
cmap=cmap_dict,
clim=oclim)
gridspace['sagittal',
mod] *= hv_ds_overlay.select(label=mod).to(
hv.Image, ['y', 'z'],
groupby=['subject_id', 'x', 'overlay_label'],
vdims='overlay',
dynamic=dynamic).opts(
cmap='glasbey_hv',
clipping_colors={
'min': 'transparent',
'NaN': 'transparent'
},
).redim.range(
overlay=(0.1, overlay_max)).apply.opts(
alpha=alpha_slider.param.value,
cmap=cmap_dict,
clim=oclim)
else:
# squish_height = int(max(image_size*(len(self.ds.z)/len(self.ds.x)), image_size/2))
# gridspace = hv.GridSpace(kdims=['label'], label=f'{self.subject_id}')
if self.verbose:
print('init gridspace')
# gridspace = hv.GridSpace(kdims=['label'])
# for mod in self.ds.label:
# gridspace[mod] = hv_ds_image.select(label=mod).to(
# hv.Image, [a1, a2], groupby=[a3], vdims='image',
# dynamic=dynamic).opts(frame_width=image_size, frame_height=image_size,
# ).apply.opts(clim=cslider.param.value)
# gridspace[mod] *= hv_ds_overlay.select(label=mod).to(
# hv.Image, [a1, a2], groupby=[a3, 'overlay_label'], vdims='overlay',
# dynamic=dynamic).opts(
# cmap='glasbey_hv', clipping_colors={'min': 'transparent'},
# ).redim.range(overlay=(1e-6, overlay_max)).apply.opts(
# alpha=alpha_slider.param.value, cmap=cmap_dict, clim=oclim)
# gridspace[mod] = gridspace[mod].opts(tools=['hover'])
# print(gridspace[mod])
gridspace = hv_ds_image.to(
hv.Image, [a1, a2],
vdims=['image', 'overlay'],
dynamic=dynamic).opts(
frame_width=pane_width,
frame_height=pane_height,
tools=['hover'],
).apply.opts(clim=cslider.param.value)
if self.verbose:
print(gridspace)
gridspace *= hv_ds_overlay.to(
hv.Image, [a1, a2], vdims='overlay', dynamic=dynamic).opts(
cmap='glasbey_hv',
clipping_colors={
'min': 'transparent',
'NaN': 'transparent'
},
).redim.range(overlay=(1e-6, overlay_max)).apply.opts(
alpha=alpha_slider.param.value,
cmap=cmap_dict,
clim=oclim)
# print(gridspace)
# print(gridspace)
# gridspace = hv.DynamicMap(subj_viewer.load_subject).grid('label')
gridspace = gridspace.layout('label')
else:
tooltips = [
('(x,y)', '($x, $y)'),
('image', '@image'),
]
hover = HoverTool(tooltips=tooltips)
hv_ds = hv.Dataset(self.ds['image'])
if self.is2d:
gridspace = hv.GridSpace(kdims=['label'])
for mod in self.ds.label.values:
gridspace[mod] = hv_ds.select(label=mod).to(
hv.Image, [a1, a2],
groupby=['subject_id'],
vdims='image',
dynamic=dynamic).opts(
frame_width=pane_width,
frame_height=pane_height,
shared_axes=False,
tools=[hover],
axiswise=True,
# ).apply.opts(clim=cslider.param.value)
)
elif three_planes:
# squish_height = int(max(image_size*(len(self.ds.z)/len(self.ds.x)), image_size/2))
# gridspace = hv.GridSpace(kdims=['plane', 'label'], label=f'{self.subject_id}')
gridspace = hv.GridSpace(kdims=['plane', 'label'])
for mod in self.ds.label.values:
gridspace['axial', mod] = hv_ds.select(label=mod).to(
hv.Image, ['x', 'y'],
groupby=['subject_id', 'z'],
vdims='image',
dynamic=dynamic).opts(frame_width=self.axial_width,
frame_height=self.axial_height,
invert_yaxis=False).apply.opts(
clim=cslider.param.value)
gridspace['coronal', mod] = hv_ds.select(label=mod).to(
hv.Image, ['x', 'z'],
groupby=['subject_id', 'y'],
vdims='image',
dynamic=dynamic).opts(
frame_width=self.coronal_width,
frame_height=self.coronal_height).apply.opts(
clim=cslider.param.value)
gridspace['sagittal', mod] = hv_ds.select(label=mod).to(
hv.Image, ['y', 'z'],
groupby=['subject_id', 'x'],
vdims='image',
dynamic=dynamic).opts(
frame_width=self.sagittal_width,
frame_height=self.sagittal_height).apply.opts(
clim=cslider.param.value)
else:
# squish_height = int(max(image_size*(len(self.ds.z)/len(self.ds.x)), image_size/2))
# gridspace = hv.GridSpace(kdims=['label'], label=f'{self.subject_id}')
gridspace = hv.GridSpace(kdims=['label'])
for mod in self.ds.label.values:
gridspace[mod] = hv_ds.select(label=mod).to(
hv.Image, [a1, a2],
groupby=['subject_id', a3],
vdims='image',
dynamic=dynamic).opts(
frame_width=pane_width,
frame_height=pane_height,
shared_axes=False,
tools=[hover],
).apply.opts(clim=cslider.param.value)
pn_layout = pn.pane.HoloViews(gridspace)
wb = pn_layout.widget_box
wb.append(cslider)
if 'overlay' in self.ds.data_vars:
wb.append(alpha_slider)
wb.append(cmap_select)
if thresh_toggle in [2, 3]:
wb.append(ocslider_max)
if thresh_toggle in [1, 3]:
wb.append(ocslider_min)
return pn.Row(wb, pn_layout)
DEFAULT_RADIUS = 1.0
DEFAULT_BOUNDS = (-DEFAULT_RADIUS, -DEFAULT_RADIUS, DEFAULT_RADIUS,
DEFAULT_RADIUS)
DEFAULT_MIN_N = 100
DEFAULT_MAX_N = 100000
DEFAULT_N_PER_PI_CALC = DEFAULT_MIN_N
DEFAULT_PLOT_SIZE = 1200
DEFAULT_IMAGE_SIZE = round(DEFAULT_PLOT_SIZE / 2)
DEFAULT_CMAP = 'Spectral'
DEFAULT_IMAGE_COLOR_IDX = 2
DEFAULT_POINT_COLOR_IDX = 125
DEFAULT_PI_UPDATE_FORMAT = 'Pi ~= {:8.7f}\nerror = {:6.3f}%\nn = {:d}\n(N ~ {:d})'
img_opts = opts.Image(cmap=DEFAULT_CMAP,
toolbar=None,
height=DEFAULT_PLOT_SIZE,
width=DEFAULT_PLOT_SIZE,
xaxis=None,
yaxis=None)
def make_circle(radius=DEFAULT_RADIUS):
def circle(t):
return (radius * np.sin(t), radius * np.cos(t), t)
lin = np.linspace(-np.pi, np.pi, 200)
return hv.Path([circle(lin)]).opts(img_opts).opts(line_width=2,
color='red')
def make_rect(bounds=DEFAULT_BOUNDS, color='blue'):
minX, minY, maxX, maxY = bounds
def create_doc(doc, data_dir):
def get_spectrogram(s0=None,
s1=None,
size=1024,
overlap=1. / 8,
zfill=1,
mode='psd'):
s_size = np.dtype(np.complex64).itemsize
if s1 is None and s0 is None:
ds = None
elif s1 is None:
ds = None
elif s0 is None:
ds = np.abs(s1)
else:
ds = np.abs(s1 - s0)
if ds is not None and s0 is not None:
flen = getsize(join(data_dir, file))
print("file size: {} bytes".format(flen))
if (ds + s0) * s_size > flen:
ds = flen - s0 * s_size
samples = np.memmap(join(data_dir, file),
dtype='complex64',
mode='r',
offset=s0 * s_size,
shape=(ds, ) if ds else None)
if ds is None:
ds = len(samples)
if ds / size > (res + 0.5) * height:
noverlap = -int(float(size) * float(ds) / size / height / res)
else:
noverlap = size // (1. / overlap)
f, t, S = signal.spectrogram(samples,
samp_rate,
nperseg=size,
nfft=int(
next_power_of_2(size) * int(zfill)),
noverlap=noverlap,
return_onesided=False,
scaling='density',
mode=mode) #
f = fftshift(f)
S = fftshift(S, axes=(0, ))
if mode == 'psd':
S = 10 * np.log10(S)
return f, t, S, samples
def get_spectrogram_img(z_min, z_max, tf_r, zfill, overlap, show_realfreq,
freq_unit, x_range, y_range):
lock.acquire()
show_realfreq = bool(show_realfreq)
hv_image = None
try:
print("y_range:", y_range, type(y_range))
if type(y_range[0]) != float:
if np.issubdtype(y_range[0],
np.datetime64) and time is not None:
y_range = [
(y - np.datetime64(time)
).astype('timedelta64[us]').astype('float') / 1e6
for y in y_range
]
#elif np.issubdtype(y0, np.timedelta64):
# y0, y1 = [y.astype('timedelta64[s]').astype('float') for y in [y0,y1]]
# tranform back to relative frequency if required
print(doc.session_context.show_realfreq, x_range)
last_freq_unit = doc.session_context.freq_unit
x_range = [x * freq_units_names[last_freq_unit] for x in x_range]
if doc.session_context.show_realfreq:
x_range = [(x - freq) for x in x_range]
print(doc.session_context.show_realfreq, "after transform",
x_range)
(x0, x1), (y0, y1) = x_range, y_range
#print("y0 dtype:", y0.dtype)
s0, s1 = sorted([
min(max(int(yr * samp_rate), 0), total_samples)
for yr in [y0, y1]
])
scale = samp_rate / np.abs(x_range[1] - x_range[0])
size = int(width *
scale) # required freq resolution to fulfill zoom level
ds = np.abs(
s1 - s0) # number of samples covered at the current zoom level
if ds / size < height:
size = int(np.sqrt(ds * scale * 10**tf_r))
f, t, S, samples = get_spectrogram(
s0,
s1,
size=size,
overlap=overlap,
mode=mode,
zfill=zfill if size < res * width else 1)
t += max(min(y0, y1), 0)
f_range = (x0 <= f) & (f <= x1)
image = S[f_range, :]
f = f[f_range]
#if ds / size > height:
# image = signal.resample(image, height*2, axis=0)
print(image.shape)
if intp_enabled:
ratio = np.array(image.shape, dtype=np.float) / np.array(
(height * res, width * res))
if np.min(ratio) < 1:
scale = np.max(
np.abs(image)
) # normalization factor for image because rescale needs that
image = rescale(image / scale, 1. / np.min(ratio),
order=1) * scale
f = signal.resample(f, image.shape[0])
t = signal.resample(t, image.shape[1])
print("after resampling: ", image.shape)
del samples
#image = hv.Image(image, bounds=(x0, y0, x1, y1)).redim.range(z=(z_min, z_max)) # TODO get exact values in bounds
if show_realtime and time is not None:
t = np.datetime64(time) + np.array(
t * 1e6).astype('timedelta64[us]')
#else:
# t = (t*1e6) #.astype('timedelta64[us]')
if show_realfreq:
f += freq
f /= freq_units_names[freq_unit]
#image = image.astype('float16') # trying to reduce network bandwidth...
print("image dtype:", image.dtype)
#hv_image = hv.Image(np.flip(image, axis=1), bounds=(min(f), max(t), max(f), min(t))) \
hv_image = hv.Image(xr.DataArray(image, coords=[f,t], dims=['f','t'], name='z'), ['f','t'], 'z') \
.options(
xlabel="Frequency [{}]".format(freq_unit),
ylabel="Time " + '[s]' if not show_realtime else '') \
.redim.range(z=(z_min, z_max))
if doc.session_context.show_realfreq != show_realfreq or doc.session_context.freq_unit != freq_unit:
hv_image = hv_image \
.redim.range(f=(min(f), max(f))) \
.redim.range(t=(min(t), max(t)))
print("redimming axis range")
doc.session_context.show_realfreq = show_realfreq
doc.session_context.freq_unit = freq_unit
except Exception as e:
print("Exception in image generation:", e)
print(traceback.format_exc())
lock.release()
return hv_image
def get_param(name, default, t=str):
try:
args = doc.session_context.request.arguments
if t == str:
p = str(args.get(name)[0], 'utf-8')
else:
p = t(args.get(name)[0])
except:
p = default
return p
time = None
freq = 0
file = basename(get_param('file', '', str))
skip = get_param('skip', 0, int)
keep = get_param('keep', None, int)
# low resolution for raspberry
width, height = get_param('width', 600, int), get_param('height', 500, int)
res = get_param('res', 1.5, float)
ds_enabled = get_param('ds', None, str) # datashade option (default: avg)
intp_enabled = get_param('intp', 0, int) # interpolation enable flap
show_realtime = bool(get_param('rt', 0,
int)) # show real time on vertical axis
mode = get_param('mode', 'psd', str)
t_range = get_param('t', None, str)
f_range = get_param('f', None, str)
if t_range is not None:
try:
parts = t_range.split(",")
if len(parts) == 2:
t_range = list(map(float, parts))
except:
pass
if f_range is not None:
try:
parts = f_range.split(",")
if len(parts) == 2:
f_range = list(map(float, parts))
elif len(parts) == 1:
_f = abs(float(parts[0]))
f_range = (-_f, _f)
except:
pass
if file.endswith(".meta"):
config = ConfigParser()
config.read(join(data_dir, file))
file = splitext(file)[0] + ".raw"
samp_rate = int(config['main']['samp_rate'])
freq = float(config['main']['freq'])
time = datetime.fromtimestamp(float(config['main']['time']))
else:
samp_rate = get_param('samp_rate', 128000, int)
f, t, S, samples = get_spectrogram(s0=skip * samp_rate,
s1=keep * samp_rate if keep else None,
size=width,
mode=mode)
total_samples = len(samples)
del samples
# default is True
if show_realtime and time is not None:
t = np.datetime64(time) + np.array(t).astype('timedelta64[s]')
doc.session_context.show_realfreq = False
doc.session_context.freq_unit = freq_units[1000]
range_stream = RangeXY(x_range=tuple(
x / freq_units_names[doc.session_context.freq_unit]
for x in ((min(f_range), max(f_range)) if f_range else (min(f),
max(f)))),
y_range=(max(t_range), min(t_range)) if t_range else
(max(t), min(t))) # transient=True
z_range = (np.min(S), np.max(S))
z_init = np.percentile(S, (50, 100))
dmap = hv.DynamicMap(
get_spectrogram_img,
streams=[range_stream],
kdims=[
hv.Dimension('z_min', range=z_range, default=z_init[0]),
hv.Dimension('z_max', range=z_range, default=z_init[1]),
hv.Dimension('tf_r',
label='Time-Frequency pixel ratio',
range=(-10., 10.),
default=0.),
hv.Dimension('zfill',
label='Zero-filling factor',
range=(1, 10),
default=2),
hv.Dimension('overlap',
label='Overlap factor',
range=(-1., 1.),
default=1. / 8),
#hv.Dimension('show_realtime', label='Show real time on vertical axis', range=(0,1), default=0),
hv.Dimension('show_realfreq',
label='Show real frequency',
range=(0, 1),
default=int(doc.session_context.show_realfreq)),
hv.Dimension('freq_unit',
label='Frequency unit',
values=list(
map(lambda x: freq_units[x],
sorted(freq_units.keys()))),
default=doc.session_context.freq_unit),
#hv.Dimension('mode', label='Spectrogram mode', values=['psd', 'angle', 'phase', 'magnitude'], default='psd')
]).options(
framewise=True, # ???
) #.redim.range(z=z_init)
#dmap = dmap.opts(opts.Image(height=height, width=width))
if ds_enabled != None:
print("datashade enabled: yes")
if ds_enabled == "" or ds_enabled == "mean":
ds_enabled = dsr.mean
elif ds_enabled == "max":
ds_enabled = dsr.max
else:
print(
"warning: invalid option for datashade. using default value: mean"
)
ds_enabled = dsr.mean
dmap = regrid(dmap,
aggregator=ds_enabled,
interpolation='linear',
upsample=True,
height=height * 2,
width=width * 2) # aggregation=dsr.max
#dmap = dmap.hist(num_bins=150, normed=False)
dmap = dmap.opts(
opts.Image(
cmap='viridis',
framewise=True,
colorbar=True,
height=height,
width=width,
tools=['hover'],
title='{}, {} {} sps'.format(
time.strftime('%Y-%m-%d %H:%M:%S') if time else 'Time unknown',
format_freq(freq), samp_rate)),
opts.Histogram(framewise=False, width=150))
#plot = renderer.get_plot(hist, doc).state
#widget = renderer.get_widget(hist, None, position='right').state
#hvobj = layout([plot, widget])
#plot = layout([renderer.get_plot(hist, doc).state])
#doc.add_root(plot)
doc = renderer.server_doc(dmap, doc=doc)
doc.title = 'Waterfall Viewer'
renderer = hv.renderer('bokeh')
# Set plot and style options
opts.defaults(
opts.Curve(xaxis=None,
yaxis=None,
show_grid=False,
show_frame=False,
color='orangered',
framewise=True,
width=100),
opts.Image(width=800,
height=400,
shared_axes=False,
logz=True,
xaxis=None,
yaxis=None,
axiswise=True), opts.HLine(color='white', line_width=1),
opts.Layout(shared_axes=False), opts.VLine(color='white', line_width=1))
# Read the parquet file
df = dd.read_parquet('./data/nyc_taxi_wide.parq').persist()
# Declare points
points = hv.Points(df, kdims=['pickup_x', 'pickup_y'], vdims=[])
# Use datashader to rasterize and linked streams for interactivity
agg = aggregate(points, link_inputs=True, x_sampling=0.0001, y_sampling=0.0001)
pointerx = hv.streams.PointerX(x=np.mean(points.range('pickup_x')),
source=points)
