# This script is part of navis (http://www.github.com/navis-org/navis).
# Copyright (C) 2018 Philipp Schlegel
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
""" Module contains functions to plot neurons in 2D/2.5D.
"""
import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.lines as mlines
import matplotlib.patches as mpatches
import matplotlib.colors as mcl
import mpl_toolkits
from mpl_toolkits.mplot3d.art3d import (Line3DCollection, Poly3DCollection,
Path3DCollection, Patch3DCollection)
from matplotlib.collections import LineCollection, PatchCollection
from matplotlib.cm import ScalarMappable
import numpy as np
import pint
import warnings
from typing import Union, List, Tuple
import copy
from typing_extensions import Literal
from .. import utils, config, core, conversion
from .colors import prepare_colormap, vertex_colors
from .plot_utils import segments_to_coords, tn_pairs_to_coords
__all__ = ['plot2d']
logger = config.get_logger(__name__)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
pint.Quantity([])
[docs]
def plot2d(x: Union[core.NeuronObject,
core.Volume,
np.ndarray,
List[Union[core.NeuronObject, np.ndarray, core.Volume]]
],
method: Union[Literal['2d'],
Literal['3d'],
Literal['3d_complex']] = '3d',
**kwargs) -> Tuple[mpl.figure.Figure, mpl.axes.Axes]:
"""Generate 2D plots of neurons and neuropils.
The main advantage of this is that you can save plot as vector graphics.
Important
---------
This function uses matplotlib which "fakes" 3D as it has only very limited
control over layering objects in 3D. Therefore neurites are not necessarily
plotted in the right Z order. This becomes especially troublesome when
plotting a complex scene with lots of neurons criss-crossing. See the
``method`` parameter for details. All methods use orthogonal projection.
Parameters
----------
x : TreeNeuron | MeshNeuron | NeuronList | Volume | Dotprops | np.ndarray
Objects to plot:
- multiple objects can be passed as list (see examples)
- numpy array of shape (n,3) is intepreted as points for
scatter plots
method : '2d' | '3d' (default) | '3d_complex'
Method used to generate plot. Comes in three flavours:
1. '2d' uses normal matplotlib. Neurons are plotted on
top of one another in the order their are passed to
the function. Use the ``view`` parameter (below) to
set the view (default = xy).
2. '3d' uses matplotlib's 3D axis. Here, matplotlib
decide the depth order (zorder) of plotting. Can
change perspective either interacively or by code
(see examples).
3. '3d_complex' same as 3d but each neuron segment is
added individually. This allows for more complex
zorders to be rendered correctly. Slows down
rendering though.
soma : bool, default=True
Plot soma if one exists. Size of the soma is determined
by the neuron's ``.soma_radius`` property which defaults
to the "radius" column for ``TreeNeurons``.
connectors : bool, default=True
Plot connectors.
connectors_only : boolean, default=False
Plot only connectors, not the neuron.
cn_size : int | float, default = 1
Size of connectors.
linewidth : int | float, default=.5
Width of neurites. Also accepts alias ``lw``.
linestyle : str, default='-'
Line style of neurites. Also accepts alias ``ls``.
autoscale : bool, default=True
If True, will scale the axes to fit the data.
scalebar : int | float | str | pint.Quantity, default=False
Adds scale bar. Provide integer, float or str to set
size of scalebar. Int|float are assumed to be in same
units as data. You can specify units in as string:
e.g. "1 um". For methods '3d' and '3d_complex', this
will create an axis object.
ax : matplotlib ax, default=None
Pass an ax object if you want to plot on an existing
canvas. Must match ``method`` - i.e. 2D or 3D axis.
figsize : tuple, default=(8, 8)
Size of figure.
color : None | str | tuple | list | dict, default=None
Use single str (e.g. ``'red'``) or ``(r, g, b)`` tuple
to give all neurons the same color. Use ``list`` of
colors to assign colors: ``['red', (1, 0, 1), ...].
Use ``dict`` to map colors to neuron IDs:
``{id: (r, g, b), ...}``.
palette : str | array | list of arrays, default=None
Name of a matplotlib or seaborn palette. If ``color`` is
not specified will pick colors from this palette.
color_by : str | array | list of arrays, default = None
Can be the name of a column in the node table of
``TreeNeurons`` or an array of (numerical or
categorical) values for each node. Numerical values will
be normalized. You can control the normalization by
passing a ``vmin`` and/or ``vmax`` parameter.
shade_by : str | array | list of arrays, default=None
Similar to ``color_by`` but will affect only the alpha
channel of the color. If ``shade_by='strahler'`` will
compute Strahler order if not already part of the node
table (TreeNeurons only). Numerical values will be
normalized. You can control the normalization by passing
a ``smin`` and/or ``smax`` parameter.
alpha : float [0-1], default=.9
Alpha value for neurons. Overriden if alpha is provided
as fourth value in ``color`` (rgb*a*). You can override
alpha value for connectors by using ``cn_alpha``.
clusters : list, default=None
A list assigning a cluster to each neuron (e.g.
``[0, 0, 0, 1, 1]``). Overrides ``color`` and uses
``palette`` to generate colors according to clusters.
depth_coloring : bool, default=False
If True, will color encode depth (Z). Overrides
``color``. Does not work with ``method = '3d_complex'``.
depth_scale : bool, default=True
If True and ``depth_coloring=True`` will plot a scale.
cn_mesh_colors : bool, default=False
If True, will use the neuron's color for its connectors.
group_neurons : bool, default=False
If True, neurons will be grouped. Works with SVG export
(not PDF). Does NOT work with ``method='3d_complex'``.
scatter_kws : dict, default={}
Parameters to be used when plotting points. Accepted
keywords are: ``size`` and ``color``.
view : tuple, default = ("x", "y")
Sets view for ``method='2d'``.
orthogonal : bool, default=True
Whether to use orthogonal or perspective view for
methods '3d' and '3d_complex'.
volume_outlines : bool | "both", default=True
If True will plot volume outline with no fill. Only
works with `method="2d"`.
dps_scale_vec : float
Scale vector for dotprops.
rasterize : bool, default=False
Neurons produce rather complex vector graphics which can
lead to large files when saving to SVG, PDF or PS. Use
this parameter to rasterize neurons and meshes/volumes
(but not axes or labels) to reduce file size.
Examples
--------
.. plot::
:context: close-figs
>>> import navis
>>> import matplotlib.pyplot as plt
Plot list of neurons as simple 2d
>>> nl = navis.example_neurons()
>>> fig, ax = navis.plot2d(nl, method='2d', view=('x', '-y'))
>>> plt.show() # doctest: +SKIP
Add a volume
.. plot::
:context: close-figs
>>> vol = navis.example_volume('LH')
>>> fig, ax = navis.plot2d([nl, vol], method='2d', view=('x', '-y'))
>>> plt.show() # doctest: +SKIP
Change neuron colors
.. plot::
:context: close-figs
>>> fig, ax = navis.plot2d(nl,
... method='2d',
... view=('x', '-y'),
... color=['r', 'g', 'b', 'm', 'c', 'y'])
>>> plt.show() # doctest: +SKIP
Plot in "fake" 3D
.. plot::
:context: close-figs
>>> fig, ax = navis.plot2d(nl, method='3d')
>>> plt.show() # doctest: +SKIP
>>> # In an interactive window you can dragging the plot to rotate
Plot in "fake" 3D and change perspective
.. plot::
:context: close-figs
>>> fig, ax = navis.plot2d(nl, method='3d')
>>> # Change view to frontal (for example neurons)
>>> ax.azim = ax.elev = 90
>>> # Change view to lateral
>>> ax.azim, ax.elev = 180, 180
>>> ax.elev = 0
>>> # Change view to top
>>> ax.azim, ax.elev = 90, 180
>>> # Tilted top view
>>> ax.azim, ax.elev = -130, -150
>>> # Move camera
>>> ax.dist = 6
>>> plt.show() # doctest: +SKIP
Plot using depth-coloring
.. plot::
:context: close-figs
>>> fig, ax = navis.plot2d(nl, method='3d', depth_coloring=True)
>>> plt.show() # doctest: +SKIP
To close all figures
>>> plt.close('all')
See the :ref:`plotting tutorial <plot_intro>` for more examples.
Returns
-------
fig, ax : matplotlib figure and axis object
See Also
--------
:func:`navis.plot3d`
Use this if you want interactive, perspectively correct renders
and if you don't need vector graphics as outputs.
:func:`navis.plot1d`
A nifty way to visualise neurons in a single dimension.
:func:`navis.plot_flat`
Plot neurons as flat structures (e.g. dendrograms).
"""
# Filter kwargs
_ACCEPTED_KWARGS = ['soma', 'connectors', 'connectors_only',
'ax', 'color', 'colors', 'c', 'view', 'scalebar',
'cn_mesh_colors', 'linewidth', 'cn_size', 'cn_alpha',
'orthogonal', 'group_neurons', 'scatter_kws', 'figsize',
'linestyle', 'rasterize', 'clusters', 'synapse_layout',
'alpha', 'depth_coloring', 'autoscale', 'depth_scale',
'ls', 'lw', 'volume_outlines', 'radius',
'dps_scale_vec', 'palette', 'color_by', 'shade_by',
'vmin', 'vmax', 'smin', 'smax', 'norm_global']
wrong_kwargs = [a for a in kwargs if a not in _ACCEPTED_KWARGS]
if wrong_kwargs:
raise KeyError(f'Unknown kwarg(s): {", ".join(wrong_kwargs)}. '
f'Currently accepted: {", ".join(_ACCEPTED_KWARGS)}')
_METHOD_OPTIONS = ['2d', '3d', '3d_complex']
if method not in _METHOD_OPTIONS:
raise ValueError(f'Unknown method "{method}". Please use either: '
f'{",".join(_METHOD_OPTIONS)}')
connectors = kwargs.get('connectors', False)
connectors_only = kwargs.get('connectors_only', False)
ax = kwargs.pop('ax', None)
scalebar = kwargs.get('scalebar', None)
# Depth coloring
depth_coloring = kwargs.get('depth_coloring', False)
depth_scale = kwargs.get('depth_scale', True)
scatter_kws = kwargs.get('scatter_kws', {})
autoscale = kwargs.get('autoscale', True)
# Parse objects
(neurons, volumes, points, _) = utils.parse_objects(x)
# Generate colors
colors = kwargs.pop('color',
kwargs.pop('c',
kwargs.pop('colors', None)))
palette = kwargs.get('palette', None)
color_by = kwargs.get('color_by', None)
shade_by = kwargs.get('shade_by', None)
# Generate the colormaps
(neuron_cmap,
volumes_cmap) = prepare_colormap(colors,
neurons=neurons,
volumes=volumes,
palette=palette,
clusters=kwargs.get('clusters', None),
alpha=kwargs.get('alpha', None),
color_range=1)
if not isinstance(color_by, type(None)):
if not palette:
raise ValueError('Must provide `palette` (e.g. "viridis") argument '
'if using `color_by`')
neuron_cmap = vertex_colors(neurons,
by=color_by,
use_alpha=False,
palette=palette,
norm_global=kwargs.get('norm_global', True),
vmin=kwargs.get('vmin', None),
vmax=kwargs.get('vmax', None),
na=kwargs.get('na', 'raise'),
color_range=1)
if not isinstance(shade_by, type(None)):
alphamap = vertex_colors(neurons,
by=shade_by,
use_alpha=True,
palette='viridis', # palette is irrelevant here
norm_global=kwargs.get('norm_global', True),
vmin=kwargs.get('smin', None),
vmax=kwargs.get('smax', None),
na=kwargs.get('na', 'raise'),
color_range=1)
new_colormap = []
for c, a in zip(neuron_cmap, alphamap):
if not (isinstance(c, np.ndarray) and c.ndim == 2):
c = np.tile(c, (a.shape[0], 1))
if c.shape[1] == 4:
c[:, 3] = a[:, 3]
else:
c = np.insert(c, 3, a[:, 3], axis=1)
new_colormap.append(c)
neuron_cmap = new_colormap
# Set axis projection
if method in ['3d', '3d_complex']:
if kwargs.get('orthogonal', True):
mpl_toolkits.mplot3d.proj3d.persp_transformation = _orthogonal_proj
else:
mpl_toolkits.mplot3d.proj3d.persp_transformation = _perspective_proj
# Generate axes
if not ax:
if method == '2d':
fig, ax = plt.subplots(figsize=kwargs.get('figsize', (8, 8)))
ax.set_aspect('equal')
elif method in ['3d', '3d_complex']:
fig = plt.figure(figsize=kwargs.get('figsize',
plt.figaspect(1) * 1.5))
ax = fig.add_subplot(111, projection='3d')
# This sets front view
ax.azim = -90
ax.elev = 0
ax.dist = 7
# Disallowed for 3D in matplotlib 3.1.0
# ax.set_aspect('equal')
# Make background transparent (nicer for dark themes)
fig.patch.set_alpha(0)
ax.patch.set_alpha(0)
# Check if correct axis were provided
else:
if not isinstance(ax, mpl.axes.Axes):
raise TypeError('Ax must be of type "mpl.axes.Axes", '
f'not "{type(ax)}"')
fig = ax.get_figure()
if method in ['3d', '3d_complex']:
if ax.name != '3d':
raise TypeError('Axis must be 3d.')
elif method == '2d':
if ax.name == '3d':
raise TypeError('Axis must be 2d.')
ax.had_data = ax.has_data()
# Prepare some stuff for depth coloring
if depth_coloring and not neurons.empty:
if method == '3d_complex':
raise Exception(f'Depth coloring unavailable for method "{method}"')
elif method == '2d':
bbox = neurons.bbox
# Add to kwargs
xy = [v.replace('-', '').replace('+', '') for v in kwargs.get('view', ('x', 'y'))]
z_ix = [v[1] for v in [('x', 0), ('y', 1), ('z', 2)] if v[0] not in xy]
kwargs['norm'] = plt.Normalize(vmin=bbox[z_ix, 0], vmax=bbox[z_ix, 1])
# Plot volumes first
if volumes:
for i, v in enumerate(volumes):
_ = _plot_volume(v,
volumes_cmap[i],
method,
ax,
**kwargs)
# Create lines from segments
visuals = {}
for i, neuron in enumerate(config.tqdm(neurons,
desc='Plot neurons',
leave=False,
disable=config.pbar_hide | len(neurons) < 2)):
if not connectors_only:
if isinstance(neuron, core.TreeNeuron) and kwargs.get('radius', False):
_neuron = conversion.tree2meshneuron(neuron)
_neuron.connectors = neuron.connectors
neuron = _neuron
if isinstance(neuron, core.TreeNeuron) and neuron.nodes.empty:
logger.warning(f'Skipping TreeNeuron w/o nodes: {neuron.label}')
if isinstance(neuron, core.TreeNeuron) and neuron.nodes.shape[0] == 1:
logger.warning(f'Skipping single-node TreeNeuron: {neuron.label}')
elif isinstance(neuron, core.MeshNeuron) and neuron.faces.size == 0:
logger.warning(f'Skipping MeshNeuron w/o faces: {neuron.label}')
elif isinstance(neuron, core.Dotprops) and neuron.points.size == 0:
logger.warning(f'Skipping Dotprops w/o points: {neuron.label}')
elif isinstance(neuron, core.TreeNeuron):
lc, sc = _plot_skeleton(neuron, neuron_cmap[i], method, ax, **kwargs)
# Keep track of visuals related to this neuron
visuals[neuron] = {'skeleton': lc, 'somata': sc}
elif isinstance(neuron, core.MeshNeuron):
m = _plot_mesh(neuron, neuron_cmap[i], method, ax, **kwargs)
visuals[neuron] = {'mesh': m}
elif isinstance(neuron, core.Dotprops):
dp = _plot_dotprops(neuron, neuron_cmap[i], method, ax, **kwargs)
visuals[neuron] = {'dotprop': dp}
elif isinstance(neuron, core.VoxelNeuron):
dp = _plot_voxels(neuron, neuron_cmap[i], method, ax, kwargs, **scatter_kws)
visuals[neuron] = {'dotprop': dp}
else:
raise TypeError(f"Don't know how to plot neuron of type '{type(neuron)}' ")
if (connectors or connectors_only) and neuron.has_connectors:
_ = _plot_connectors(neuron, neuron_cmap[i], method, ax, **kwargs)
for p in points:
_ = _plot_scatter(p, method, ax, kwargs, **scatter_kws)
if autoscale:
if method == '2d':
ax.autoscale(tight=True)
elif method in ['3d', '3d_complex']:
# Make sure data lims are set correctly
update_axes3d_bounds(ax)
# Rezie to have equal aspect
set_axes3d_equal(ax)
if scalebar is not None:
_ = _add_scalebar(scalebar, neurons, method, ax)
def set_depth():
"""Set depth information for neurons according to camera position."""
# Get projected coordinates
proj_co = mpl_toolkits.mplot3d.proj3d.proj_points(all_co, ax.get_proj())
# Get min and max of z coordinates
z_min, z_max = min(proj_co[:, 2]), max(proj_co[:, 2])
# Generate a new normaliser
norm = plt.Normalize(vmin=z_min, vmax=z_max)
# Go over all neurons and update Z information
for neuron in visuals:
# Get this neurons colletion and coordinates
if 'skeleton' in visuals[neuron]:
c = visuals[neuron]['skeleton']
this_co = c._segments3d[:, 0, :]
elif 'mesh' in visuals[neuron]:
c = visuals[neuron]['mesh']
# Note that we only get every third position -> that's because
# these vectors actually represent faces, i.e. each vertex
this_co = c._vec.T[::3, [0, 1, 2]]
else:
raise ValueError(f'Neither mesh nor skeleton found for neuron {neuron.id}')
# Get projected coordinates
this_proj = mpl_toolkits.mplot3d.proj3d.proj_points(this_co,
ax.get_proj())
# Normalise z coordinates
ns = norm(this_proj[:, 2]).data
# Set array
c.set_array(ns)
# No need for normaliser - already happened
c.set_norm(None)
if (isinstance(neuron, core.TreeNeuron)
and not isinstance(getattr(neuron, 'soma', None), type(None))):
# Get depth of soma(s)
soma = utils.make_iterable(neuron.soma)
soma_co = neuron.nodes.set_index('node_id').loc[soma][['x', 'y', 'z']].values
soma_proj = mpl_toolkits.mplot3d.proj3d.proj_points(soma_co,
ax.get_proj())
soma_cs = norm(soma_proj[:, 2]).data
# Set soma color
for cs, s in zip(soma_cs, visuals[neuron]['somata']):
s.set_color(cmap(cs))
def Update(event):
set_depth()
if depth_coloring:
cmap = mpl.cm.jet
if method == '2d' and depth_scale:
sm = ScalarMappable(norm=kwargs['norm'], cmap=cmap)
fig.colorbar(sm, ax=ax, fraction=.075, shrink=.5, label='Depth')
elif method == '3d':
# Collect all coordinates
all_co = []
for n in visuals:
if 'skeleton' in visuals[n]:
all_co.append(visuals[n]['skeleton']._segments3d[:, 0, :])
if 'mesh' in visuals[n]:
all_co.append(visuals[n]['mesh']._vec.T[:, [0, 1, 2]])
all_co = np.concatenate(all_co, axis=0)
fig.canvas.mpl_connect('draw_event', Update)
set_depth()
plt.axis('off')
return fig, ax
def _add_scalebar(scalebar, neurons, method, ax):
"""Add scalebar."""
if isinstance(scalebar, bool):
scalebar = '1 um'
if isinstance(scalebar, str):
scalebar = config.ureg(scalebar)
if isinstance(scalebar, pint.Quantity):
# If we have neurons as points of reference convert
if neurons:
scalebar = scalebar.to(neurons[0].units).magnitude
# If no reference, use assume it's the same units
else:
scalebar = scalebar.magnitude
# Hard-coded offset from figure boundaries
ax_offset = (ax.get_xlim()[1] - ax.get_xlim()[0]) / 100 * 5
if method == '2d':
xlim = ax.get_xlim()
ylim = ax.get_ylim()
coords = np.array([[xlim[0] + ax_offset, ylim[0] + ax_offset],
[xlim[0] + ax_offset + scalebar, ylim[0] + ax_offset]
])
sbar = mlines.Line2D(
coords[:, 0], coords[:, 1], lw=3, alpha=.9, color='black')
sbar.set_gid(f'{scalebar}_scalebar')
ax.add_line(sbar)
elif method in ['3d', '3d_complex']:
xlim = ax.get_xlim()
ylim = ax.get_ylim()
zlim = ax.get_zlim()
left = xlim[0] + ax_offset
bottom = zlim[0] + ax_offset
front = ylim[0] + ax_offset
sbar = [np.array([[left, front, bottom],
[left, front, bottom]]),
np.array([[left, front, bottom],
[left, front, bottom]]),
np.array([[left, front, bottom],
[left, front, bottom]])]
sbar[0][1][0] += scalebar
sbar[1][1][1] += scalebar
sbar[2][1][2] += scalebar
lc = Line3DCollection(sbar, color='black', lw=1)
lc.set_gid(f'{scalebar}_scalebar')
ax.add_collection3d(lc)
def _plot_scatter(points, method, ax, kwargs, **scatter_kws):
"""Plot dotprops."""
if method == '2d':
default_settings = dict(
c='black',
zorder=4,
edgecolor='none',
s=1
)
default_settings.update(scatter_kws)
default_settings = _fix_default_dict(default_settings)
view = kwargs.get('view', ('x', 'y'))
x, y = _parse_view2d(points, view)
ax.scatter(x, y, **default_settings)
elif method in ['3d', '3d_complex']:
default_settings = dict(
c='black',
s=1,
depthshade=False,
edgecolor='none'
)
default_settings.update(scatter_kws)
default_settings = _fix_default_dict(default_settings)
ax.scatter(points[:, 0],
points[:, 1],
points[:, 2],
**default_settings
)
def _plot_voxels(vx, color, method, ax, kwargs, **scatter_kws):
"""Plot VoxelNeuron as scatter plot."""
# Use only the top N voxels
assert isinstance(vx, core.VoxelNeuron)
n_pts = 1000000
v = vx.values
pts = vx.voxels
srt = np.argsort(v)[::-1]
pts = pts[srt][: n_pts]
v = v[srt][: n_pts]
# Scale points by units
pts = pts * vx.units_xyz.magnitude + vx.offset
# Calculate colors
cmap = color_to_cmap(color)
colors = cmap(v / v.max())
if method == '2d':
view = kwargs.get('view', ('x', 'y'))
x, y = _parse_view2d(pts, view)
ax.scatter(x, y,
c=colors,
s=scatter_kws.get('size', 20))
elif method in ['3d', '3d_complex']:
ax.scatter(pts[:, 0], pts[:, 1], pts[:, 2],
c=colors,
marker=scatter_kws.get('marker', 'o'),
s=scatter_kws.get('size', .1))
def color_to_cmap(color):
"""Convert single color to color palette."""
color = mcl.to_rgb(color)
colors = [[color[0], color[1], color[2], 0],
[color[0], color[1], color[2], 1]]
return mcl.LinearSegmentedColormap.from_list('Palette', colors, N=256)
def _plot_dotprops(dp, color, method, ax, **kwargs):
"""Plot dotprops."""
# Here, we will effectively cheat and turn the dotprops into a skeleton
# which we can then pass to _plot_skeleton
tn = dp.to_skeleton(scale_vec=kwargs.get('dps_scale_vec', 'auto'))
return _plot_skeleton(tn, color, method, ax, **kwargs)
def _plot_connectors(neuron, color, method, ax, **kwargs):
"""Plot connectors."""
view = kwargs.get('view', ('x', 'y'))
if not kwargs.get('cn_mesh_colors', False):
cn_layout = config.default_connector_colors.copy()
else:
cn_layout = copy.deepcopy(config.default_connector_colors)
# change all of the colors to color
for inner_dict in cn_layout.values():
if not isinstance(inner_dict, dict):
continue
inner_dict["color"] = color
cn_layout.update(kwargs.get('synapse_layout', {}))
if method == '2d':
for c in neuron.connectors.type.unique():
this_cn = neuron.connectors[neuron.connectors.type == c]
x, y = _parse_view2d(this_cn[['x', 'y', 'z']].values, view)
ax.scatter(x, y,
color=cn_layout[c]['color'],
edgecolor='none',
s=kwargs.get('cn_size', cn_layout['size']))
ax.get_children()[-1].set_gid(f'CN_{neuron.id}')
elif method in ['3d', '3d_complex']:
all_cn = neuron.connectors
c = [cn_layout[i]['color'] for i in all_cn.type.values]
ax.scatter(all_cn.x.values, all_cn.y.values, all_cn.z.values,
color=c,
s=kwargs.get('cn_size', cn_layout['size']),
depthshade=cn_layout.get('depthshade', False),
edgecolor='none')
ax.get_children()[-1].set_gid(f'CN_{neuron.id}')
def _plot_mesh(neuron, color, method, ax, **kwargs):
"""Plot mesh (i.e. MeshNeuron)."""
name = getattr(neuron, 'name')
depth_coloring = kwargs.get('depth_coloring', False)
alpha = kwargs.get('alpha', None)
group_neurons = kwargs.get('group_neurons', False)
view = kwargs.get('view', ('x', 'y'))
rasterize = kwargs.get('rasterize', False)
# Add alpha
if alpha:
color = (color[0], color[1], color[2], alpha)
ts = None
if method == '2d':
# Generate 2d representation
xy = np.dstack(_parse_view2d(neuron.vertices, view))[0]
# Map vertex colors to faces
if isinstance(color, np.ndarray) and color.ndim == 2:
if len(color) != len(neuron.faces) and len(color) == len(neuron.vertices):
color = [color[f].mean(axis=0)[:3].tolist() for f in neuron.faces]
# Generate a patch for each face
patches = []
for i, f in enumerate(neuron.faces):
p = mpatches.Polygon(xy[f], closed=True)
patches.append(p)
pc = PatchCollection(patches, linewidth=0, facecolor=color,
rasterized=rasterize,
edgecolor='none', alpha=alpha)
ax.add_collection(pc)
else:
ts = ax.plot_trisurf(neuron.vertices[:, 0],
neuron.vertices[:, 1],
neuron.faces,
neuron.vertices[:, 2],
label=name,
rasterized=rasterize,
cmap=mpl.cm.jet if depth_coloring else None,
color=color)
if group_neurons:
ts.set_gid(neuron.id)
return ts
def _get_depth_axis(view):
"""Return index of axis which is not used for x/y."""
view = [v.replace('-', '').replace('+', '') for v in view]
depth = [ax for ax in ['x', 'y', 'z']][0]
map = {'x': 0, 'y': 1, 'z': 2}
return map[depth]
def _parse_view2d(co, view):
"""Parse view parameter and returns x/y parameter."""
if not isinstance(co, np.ndarray):
co = np.array(co)
map = {'x': 0, 'y': 1, 'z': 2}
x_ix = map[view[0].replace('-', '').replace('+', '')]
y_ix = map[view[1].replace('-', '').replace('+', '')]
x_mod = -1 if '-' in view[0] else 1
y_mod = -1 if '-' in view[1] else 1
if co.ndim == 2:
x = co[:, x_ix]
y = co[:, y_ix]
# Multiply only where co is not None
x = np.multiply(x, x_mod, where=x != None, subok=False)
y = np.multiply(y, y_mod, where=y != None, subok=False)
# Do NOT remove the list() here - for some reason the multiplication
# above causes issues in matplotlib
return (list(x), list(y))
elif co.ndim == 3:
xy = co[:, :, [x_ix, y_ix]] * [x_mod, y_mod]
return xy
else:
raise ValueError(f'Expect coordinates to have 2 or 3 dimensions, got {co.ndim}')
def _plot_skeleton(neuron, color, method, ax, **kwargs):
"""Plot skeleton."""
depth_coloring = kwargs.get('depth_coloring', False)
linewidth = kwargs.get('linewidth', kwargs.get('lw', .5))
linestyle = kwargs.get('linestyle', kwargs.get('ls', '-'))
alpha = kwargs.get('alpha', None)
norm = kwargs.get('norm')
plot_soma = kwargs.get('soma', True)
group_neurons = kwargs.get('group_neurons', False)
view = kwargs.get('view', ('x', 'y'))
rasterize = kwargs.get('rasterize', False)
if method == '2d':
if not depth_coloring and not (isinstance(color, np.ndarray) and color.ndim == 2):
# Generate by-segment coordinates
coords = segments_to_coords(neuron,
neuron.segments,
modifier=(1, 1, 1))
# We have to add (None, None, None) to the end of each
# slab to make that line discontinuous there
coords = np.vstack([np.append(t, [[None] * 3],
axis=0) for t in coords])
x, y = _parse_view2d(coords, view)
this_line = mlines.Line2D(x, y,
lw=linewidth, ls=linestyle,
alpha=alpha, color=color,
rasterized=rasterize,
label=f'{getattr(neuron, "name", "NA")} - #{neuron.id}')
ax.add_line(this_line)
else:
coords = tn_pairs_to_coords(neuron, modifier=(1, 1, 1))
xy = _parse_view2d(coords, view)
lc = LineCollection(xy,
cmap='jet' if depth_coloring else None,
norm=norm if depth_coloring else None,
rasterized=rasterize,
joinstyle='round')
lc.set_linewidth(linewidth)
lc.set_linestyle(linestyle)
lc.set_label(f'{getattr(neuron, "name", "NA")} - #{neuron.id}')
if depth_coloring:
lc.set_alpha(alpha)
lc.set_array(neuron.nodes.loc[neuron.nodes.parent_id >= 0, 'z'].values)
elif (isinstance(color, np.ndarray) and color.ndim == 2):
# If we have a color for each node, we need to drop the roots
if color.shape[1] != coords.shape[0]:
lc.set_color(color[neuron.nodes.parent_id.values >= 0])
else:
lc.set_color(color)
ax.add_collection(lc)
if plot_soma and np.any(neuron.soma):
soma = utils.make_iterable(neuron.soma)
# If soma detection is messed up we might end up producing
# dozens of soma which will freeze the kernel
if len(soma) >= 10:
logger.warning(f'{neuron.id} - {len(soma)} somas found.')
for s in soma:
if isinstance(color, np.ndarray) and color.ndim > 1:
s_ix = np.where(neuron.nodes.node_id == s)[0][0]
soma_color = color[s_ix]
else:
soma_color = color
n = neuron.nodes.set_index('node_id').loc[s]
r = getattr(n, neuron.soma_radius) if isinstance(neuron.soma_radius, str) else neuron.soma_radius
if depth_coloring:
d = [n.x, n.y, n.z][_get_depth_axis(view)]
soma_color = mpl.cm.jet(norm(d))
sx, sy = _parse_view2d(np.array([[n.x, n.y, n.z]]), view)
c = mpatches.Circle((sx[0], sy[0]), radius=r,
alpha=alpha, fill=True, fc=soma_color,
rasterized=rasterize,
zorder=4, edgecolor='none')
ax.add_patch(c)
return None, None
elif method in ['3d', '3d_complex']:
# For simple scenes, add whole neurons at a time to speed up rendering
if method == '3d':
if (isinstance(color, np.ndarray) and color.ndim == 2) or depth_coloring:
coords = tn_pairs_to_coords(neuron,
modifier=(1, 1, 1))
# If we have a color for each node, we need to drop the roots
if isinstance(color, np.ndarray) and color.shape[1] != coords.shape[0]:
line_color = color[neuron.nodes.parent_id.values >= 0]
else:
line_color = color
else:
# Generate by-segment coordinates
coords = segments_to_coords(neuron,
neuron.segments,
modifier=(1, 1, 1))
line_color = color
lc = Line3DCollection(coords,
color=line_color if not depth_coloring else None,
label=neuron.id,
alpha=alpha,
cmap=None if not depth_coloring else mpl.cm.jet,
lw=linewidth,
joinstyle='round',
rasterized=rasterize,
linestyle=linestyle)
if group_neurons:
lc.set_gid(neuron.id)
# Need to get this before adding data
line3D_collection = lc
ax.add_collection3d(lc)
# For complex scenes, add each segment as a single collection
# -> helps reducing Z-order errors
elif method == '3d_complex':
# Generate by-segment coordinates
coords = segments_to_coords(neuron,
neuron.segments,
modifier=(1, 1, 1))
for c in coords:
lc = Line3DCollection([c],
color=color,
lw=linewidth,
alpha=alpha,
rasterized=rasterize,
linestyle=linestyle)
if group_neurons:
lc.set_gid(neuron.id)
ax.add_collection3d(lc)
line3D_collection = None
surf3D_collections = []
if plot_soma and not isinstance(getattr(neuron, 'soma', None),
type(None)):
soma = utils.make_iterable(neuron.soma)
# If soma detection is messed up we might end up producing
# dozens of soma which will freeze the kernel
if len(soma) >= 5:
logger.warning(f'Neuron {neuron.id} appears to have {len(soma)}'
' somas. Skipping plotting its somas.')
else:
for s in soma:
if isinstance(color, np.ndarray) and color.ndim > 1:
s_ix = np.where(neuron.nodes.node_id == s)[0][0]
soma_color = color[s_ix]
else:
soma_color = color
n = neuron.nodes.set_index('node_id').loc[s]
r = getattr(n, neuron.soma_radius) if isinstance(neuron.soma_radius, str) else neuron.soma_radius
resolution = 20
u = np.linspace(0, 2 * np.pi, resolution)
v = np.linspace(0, np.pi, resolution)
x = r * np.outer(np.cos(u), np.sin(v)) + n.x
y = r * np.outer(np.sin(u), np.sin(v)) + n.y
z = r * np.outer(np.ones(np.size(u)), np.cos(v)) + n.z
surf = ax.plot_surface(x, y, z,
color=soma_color,
shade=False,
rasterized=rasterize,
alpha=alpha)
if group_neurons:
surf.set_gid(neuron.id)
surf3D_collections.append(surf)
return line3D_collection, surf3D_collections
def _plot_volume(volume, color, method, ax, **kwargs):
"""Plot volume."""
name = getattr(volume, 'name')
rasterize = kwargs.get('rasterize', False)
if len(color) == 4:
this_alpha = color[3]
else:
this_alpha = 1
if kwargs.get('volume_outlines', False):
fill, lw, fc, ec = False, 1, 'none', color
else:
fill, lw, fc, ec = True, 0, color, 'none'
if method == '2d':
view = kwargs.get('view', ('x', 'y'))
volume_outlines = kwargs.get('volume_outlines', False)
if volume_outlines in (False, 'both'):
# Generate 2d representation
xy = np.dstack(_parse_view2d(volume.verts, view))[0]
# Generate a patch for each face
patches = []
for f in volume.faces:
p = mpatches.Polygon(xy[f], closed=True, fill=fill)
patches.append(p)
pc = PatchCollection(patches, linewidth=lw, facecolor=fc,
rasterized=rasterize,
edgecolor=ec, alpha=this_alpha, zorder=0)
ax.add_collection(pc)
if volume_outlines in (True, 'both'):
verts = volume.to_2d(view=view, alpha=0.001)
vpatch = mpatches.Polygon(verts, closed=True, lw=lw, fill=fill,
rasterized=rasterize,
fc=fc, ec=ec, zorder=0,
alpha=1 if volume_outlines == 'both' else this_alpha)
ax.add_patch(vpatch)
elif method in ['3d', '3d_complex']:
verts = np.vstack(volume.vertices)
# Add alpha
if len(color) == 3:
color = (color[0], color[1], color[2], .1)
ts = ax.plot_trisurf(verts[:, 0],
verts[:, 1],
volume.faces,
verts[:, 2],
label=name,
rasterized=rasterize,
color=color)
ts.set_gid(name)
def update_axes3d_bounds(ax):
"""Update axis bounds and remove default points (0,0,0) and (1,1,1)."""
# Collect data points present in the figure
points = []
for c in ax.collections:
if isinstance(c, Line3DCollection):
for s in c._segments3d:
points.append(s)
elif isinstance(c, Poly3DCollection):
points.append(c._vec[:3, :].T)
elif isinstance(c, (Path3DCollection, Patch3DCollection)):
points.append(np.array(c._offsets3d).T)
if not len(points):
return
points = np.vstack(points)
# If this is the first set of points, we need to overwrite the defaults
# That should happen automatically but for some reason doesn't for 3d axes
if not getattr(ax, 'had_data', False):
mn = points.min(axis=0)
mx = points.max(axis=0)
new_xybounds = np.array([[mn[0], mn[1]],
[mx[0], mx[1]]])
new_zzbounds = np.array([[mn[2], mn[2]],
[mx[2], mx[2]]])
ax.xy_dataLim.set_points(new_xybounds)
ax.zz_dataLim.set_points(new_zzbounds)
ax.xy_viewLim.set_points(new_xybounds)
ax.zz_viewLim.set_points(new_zzbounds)
ax.had_data = True
else:
ax.auto_scale_xyz(points[:, 0].tolist(),
points[:, 1].tolist(),
points[:, 2].tolist(),
had_data=True)
def set_axes3d_equal(ax):
"""Make axes of 3D plot have equal scale.
This requires the viewLim to be set correctly: see `update_axes3d_bounds()`.
Parameters
----------
ax : a matplotlib axis, e.g., as output from plt.gca().
"""
x_limits = ax.get_xlim3d()
y_limits = ax.get_ylim3d()
z_limits = ax.get_zlim3d()
x_range = abs(x_limits[1] - x_limits[0])
x_middle = np.mean(x_limits)
y_range = abs(y_limits[1] - y_limits[0])
y_middle = np.mean(y_limits)
z_range = abs(z_limits[1] - z_limits[0])
z_middle = np.mean(z_limits)
# The plot bounding box is a sphere in the sense of the infinity
# norm, hence I call half the max range the plot radius.
plot_radius = 0.5*max([x_range, y_range, z_range])
ax.set_xlim3d([x_middle - plot_radius, x_middle + plot_radius])
ax.set_ylim3d([y_middle - plot_radius, y_middle + plot_radius])
ax.set_zlim3d([z_middle - plot_radius, z_middle + plot_radius])
# Set 1:1:1 box ratio
ax.set_box_aspect((1, 1, 1))
def __old__update_axes3d_bounds(ax, points):
"""Update axis bounds and remove default points (0,0,0) and (1,1,1)."""
if not isinstance(points, np.ndarray):
points = np.ndarray(points)
# If this is the first set of points, we need to overwrite the defaults
# That should happen automatically but for some reason doesn't for 3d axes
if not getattr(ax, 'had_data', False):
mn = points.min(axis=0)
mx = points.max(axis=0)
new_xybounds = np.array([[mn[0], mn[1]],
[mx[0], mx[1]]])
new_zzbounds = np.array([[mn[2], mn[2]],
[mx[2], mx[2]]])
ax.xy_dataLim.set_points(new_xybounds)
ax.zz_dataLim.set_points(new_zzbounds)
ax.had_data = True
else:
ax.auto_scale_xyz(points[:, 0].tolist(),
points[:, 1].tolist(),
points[:, 2].tolist(),
had_data=True)
def _fix_default_dict(x):
"""Consolidate duplicate settings.
E.g. scatter kwargs when 'c' and 'color' is provided.
"""
# The first entry is the "survivor"
duplicates = [['color', 'c'], ['size', 's'], ['alpha', 'a']]
for dupl in duplicates:
if sum([v in x for v in dupl]) > 1:
to_delete = [v for v in dupl if v in x][1:]
_ = [x.pop(v) for v in to_delete]
return x
def _perspective_proj(zfront, zback, focal_length=1):
"""Copy of the original matplotlib projection matrix.
Notably, we set a default value for focal_length because this was only added
with version 3.6 of matplotlib.
"""
e = focal_length
a = 1 # aspect ratio
b = (zfront+zback)/(zfront-zback)
c = -2*(zfront*zback)/(zfront-zback)
proj_matrix = np.array([[e, 0, 0, 0],
[0, e/a, 0, 0],
[0, 0, b, c],
[0, 0, -1, 0]])
return proj_matrix
def _orthogonal_proj(zfront, zback, focal_length=None):
"""Get matplotlib to use orthogonal instead of perspective view.
Usage:
proj3d.persp_transformation = _orthogonal_proj
"""
a = (zfront + zback) / (zfront - zback)
b = -2 * (zfront * zback) / (zfront - zback)
# -0.0001 added for numerical stability as suggested in:
# http://stackoverflow.com/questions/23840756
return np.array([[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, a, b],
[0, 0, -0.0001, zback]])