I have computed a lot (~5000) of 3d points (x,y,z) in a quite complicated way so I have no function such that z = f(x,y). I can plot the 3d surface using
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm
X = surface_points[:,0]
Y = surface_points[:,1]
Z = surface_points[:,2]
fig = plt.figure()
ax = fig.add_subplot(projection='3d')
surf = ax.plot_trisurf(X, Y, Z, cmap=cm.coolwarm, vmin=np.nanmin(Z), vmax=np.nanmax(Z))
I would like to plot this also in 2d, with a colorbar indicating the z-value. I know there is a simple solution using ax.contour if my z is a matrix, but here I only have a vector.
Attaching the plot_trisurf result when rotated to xy-plane. This is what I what like to achieve without having to rotate a 3d plot. In this, my variable surface_points is an np.array with size 5024 x 3.
I had the same problems in one of my codes, I solved it this way:
import numpy as np
from scipy.interpolate import griddata
import matplotlib.pylab as plt
from matplotlib import cm
N = 10000
surface_points = np.random.rand(N,3)
X = surface_points[:,0]
Y = surface_points[:,1]
Z = surface_points[:,2]
nx = 10*int(np.sqrt(N))
xg = np.linspace(X.min(), X.max(), nx)
yg = np.linspace(Y.min(), Y.max(), nx)
xgrid, ygrid = np.meshgrid(xg, yg)
ctr_f = griddata((X, Y), Z, (xgrid, ygrid), method='linear')
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.contourf(xgrid, ygrid, ctr_f, cmap=cm.coolwarm)
plt.show()
You could use a scatter plot to display a projection of your z color onto the x-y axis.
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm
N = 10000
surface_points = np.random.rand(N,3)
X = surface_points[:,0]
Y = surface_points[:,1]
Z = surface_points[:,2]
# fig = plt.figure()
# ax = fig.add_subplot(projection='3d')
# surf = ax.plot_trisurf(X, Y, Z, cmap=cm.coolwarm, vmin=np.nanmin(Z), vmax=np.nanmax(Z))
fig = plt.figure()
cmap = cm.get_cmap('coolwarm')
color = cmap(Z)[..., :3]
plt.scatter(X,Y,c=color)
plt.show()
Since you seem to have a 3D shape that is hollow, you could split the projection into two like if you cur the shape in two pieces.
fig = plt.figure()
plt.subplot(121)
plt.scatter(X[Z<0.5],Y[Z<0.5],c=color[Z<0.5])
plt.title('down part')
plt.subplot(122)
plt.scatter(X[Z>=0.5],Y[Z>=0.5],c=color[Z>+0.5])
plt.title('top part')
plt.show()
I'm working on a 3D plot displayed by a wireframe, where 2D plots are projected on the x, y, and z surface, respectively. Below you can find a minimum example.
I have 2 questions:
With contourf, the 2D plots for every x=10, x=20,... or y=10, y=20,... are displayed on the plot walls. Is there a possibility to define for which x or y, respectively, the contour plots are displayed? For example, in case I only want to have the xz contour plot for y = 0.5 mirrored on the wall?
ADDITION: To display what I mean with "2D plots", I changed "contourf" in the code to "contour" and added the resulting plot to this question. Here you can see now the xz lines for different y values, all offset to y=90. What if I do not want to have all the lines, but only two of them for defined y values?
3D_plot_with_2D_contours
As you can see in the minimum example, the 2D contour plot optically covers the wireframe 3D plot. With increasing the transparency with alpha=0.5 I can increase the transparency of the 2D contours to at least see the wireframe, but it is still optically wrong. Is it possible to sort the objects correctly?
import matplotlib.pyplot as plt,numpy as np
import pylab as pl
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt,numpy as np
plt.clf()
fig = plt.figure(1,figsize=(35,17),dpi=600,facecolor='w',edgecolor='k')
fig.set_size_inches(10.5,8)
ax = fig.gca(projection='3d')
X, Y, Z = axes3d.get_test_data(0.05)
Xnew = X + 50
Ynew = Y + 50
cset = ax.contourf(Xnew, Ynew, Z, zdir='z', offset=-100, cmap=plt.cm.coolwarm, alpha=0.5)
cset = ax.contourf(Xnew, Ynew, Z, zdir='x', offset=10, cmap=plt.cm.coolwarm, alpha=0.5)
cset = ax.contourf(Xnew, Ynew, Z, zdir='y', offset=90, cmap=plt.cm.coolwarm, alpha = 0.5)
ax.plot_wireframe(Xnew, Ynew, Z, rstride=5, cstride=5, color='black')
Z=Z-Z.min()
Z=Z/Z.max()
from scipy.ndimage.interpolation import zoom
Xall=zoom(Xnew,5)
Yall=zoom(Ynew,5)
Z=zoom(Z,5)
ax.set_xlim(10, 90)
ax.set_ylim(10, 90)
ax.set_zlim(-100, 100)
ax.tick_params(axis='z', which='major', pad=10)
ax.set_xlabel('X',labelpad=10)
ax.set_ylabel('Y',labelpad=10)
ax.set_zlabel('Z',labelpad=17)
ax.view_init(elev=35., azim=-70)
fig.tight_layout()
plt.show()
ADDITION 2: Here is the actual code I'm working with. However, the original data are hidden in the csv files which are too big to be included in the minimal example. That's why was initially replacing them by the test data. However, maybe the actual code helps nevertheless.
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt,numpy as np
import pylab as pl
from matplotlib.markers import MarkerStyle
import csv
with open("X.csv", 'r') as f:
X = list(csv.reader(f, delimiter=";"))
import numpy as np
X = np.array(X[1:], dtype=np.float)
import csv
with open("Z.csv", 'r') as f:
Z = list(csv.reader(f, delimiter=";"))
import numpy as np
Z = np.array(Z[1:], dtype=np.float)
Y = [[7,7.1,7.2,7.3,7.4,7.5,7.6,7.7,7.8,7.9,8,8.1,8.2,8.3,8.4,8.5,8.6,8.7,8.8,8.9,9]]
Xall = np.repeat(X[:],21,axis=1)
Yall = np.repeat(Y[:],30,axis=0)
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt,numpy as np
plt.clf()
fig = plt.figure(1,figsize=(35,17),dpi=600,facecolor='w',edgecolor='k')
fig.set_size_inches(10.5,8)
ax = fig.gca(projection='3d')
cset = ax.contourf(Xall, Yall, Z, 2, zdir='x', offset=0, cmap=plt.cm.coolwarm, shade = False, edgecolor='none', alpha=0.5)
cset = ax.contourf(Xall, Yall, Z, 2, zdir='y', offset=9, cmap=plt.cm.coolwarm, shade = False, edgecolor='none', alpha=0.5)
ax.plot_wireframe(Xall, Yall, Z, rstride=1, cstride=1, color='black')
Z=Z-Z.min()
Z=Z/Z.max()
from scipy.ndimage.interpolation import zoom
Xall=zoom(Xall,5)
Yall=zoom(Yall,5)
Z=zoom(Z,5)
cset = ax.plot_surface(Xall, Yall, np.zeros_like(Z)-0,facecolors=plt.cm.coolwarm(Z),shade=False,alpha=0.5,linewidth=False)
ax.set_xlim(-0.5, 31)
ax.set_ylim(6.9, 9.1)
ax.set_zlim(0, 500)
labelsx = [item.get_text() for item in ax.get_xticklabels()]
empty_string_labelsx = ['']*len(labelsx)
ax.set_xticklabels(empty_string_labelsx)
labelsy = [item.get_text() for item in ax.get_yticklabels()]
empty_string_labelsy = ['']*len(labelsy)
ax.set_yticklabels(empty_string_labelsy)
labelsz = [item.get_text() for item in ax.get_zticklabels()]
empty_string_labelsz = ['']*len(labelsz)
ax.set_zticklabels(empty_string_labelsz)
import matplotlib.ticker as ticker
ax.xaxis.set_major_locator(ticker.MultipleLocator(5))
ax.xaxis.set_minor_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(0.5))
ax.yaxis.set_minor_locator(ticker.MultipleLocator(0.25))
ax.zaxis.set_major_locator(ticker.MultipleLocator(100))
ax.zaxis.set_minor_locator(ticker.MultipleLocator(50))
ax.tick_params(axis='z', which='major', pad=10)
ax.set_xlabel('X',labelpad=5,fontsize=15)
ax.set_ylabel('Y',labelpad=5,fontsize=15)
ax.set_zlabel('Z',labelpad=5,fontsize=15)
ax.view_init(elev=35., azim=-70)
fig.tight_layout()
plt.show()
Alternate possible answer.
This code demonstrates
A plot of a surface and its correponding wireframe
The creation of data and its plot of 3d lines (draped on the surface in 1) at specified values of x and y
Projections of the 3d lines (in 2) on to the frame walls
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt
from scipy import interpolate
import numpy as np
# use the test data for plotting
fig = plt.figure(1, figsize=(6,6), facecolor='w', edgecolor='gray')
ax = fig.gca(projection='3d')
X, Y, Z = axes3d.get_test_data(0.1) #get 3d data at appropriate density
# create an interpolating function
# can take a long time if data is too large
f1 = interpolate.interp2d(X, Y, Z, kind='linear')
# in general, one can use a set of other X,Y,Z that cover a surface
# preferably, (X,Y) are in grid arrangement
# make up a new set of 3d data to plot
# ranges of x1, and y1 will be inside (X,Y) of the data obtained above
# related grid, x1g,y1g,z1g will be obtained from meshgrid and the interpolated function
x1 = np.linspace(-15,15,10)
y1 = np.linspace(-15,15,10)
x1g, y1g = np.meshgrid(x1, y1)
z1g = f1(x1, y1) #dont use (x1g, y1g)
# prep data for 3d line on the surface (X,Y,Z) at x=7.5
n = 12
x_pf = 7.5
x5 = x_pf*np.ones(n)
y5 = np.linspace(-15, 15, n)
z5 = f1(x_pf, y5)
# x5,y5,z5 can be used to plot 3d line on the surface (X,Y,Z)
# prep data for 3d line on the surface (X,Y,Z) at y=6
y_pf = 6
x6 = np.linspace(-15, 15, n)
y6 = x_pf*np.ones(n)
z6 = f1(x6, y_pf)
# x6,y6,z6 can be used to plot 3d line on the surface (X,Y,Z)
ax = fig.gca(projection='3d')
ax.plot_surface(x1g, y1g, z1g, alpha=0.25)
ax.plot_wireframe(x1g, y1g, z1g, rstride=2, cstride=2, color='black', zorder=10, alpha=1, lw=0.8)
# 3D lines that follow the surface
ax.plot(x5,y5,z5.flatten(), color='red', lw=4)
ax.plot(x6,y6,z6.flatten(), color='green', lw=4)
# projections of 3d curves
# project red and green lines to the walls
ax.plot(-15*np.ones(len(y5)), y5, z5.flatten(), color='red', lw=4, linestyle=':', alpha=0.6)
ax.plot(x6, 15*np.ones(len(x6)), z6.flatten(), color='green', lw=4, linestyle=':', alpha=0.6)
# projections on other sides (become vertical lines)
# change to if True, to plot these
if False:
ax.plot(x5, 15*np.ones(len(x5)), z5.flatten(), color='red', lw=4, alpha=0.3)
ax.plot(-15*np.ones(len(x6)), y6, z6.flatten(), color='green', lw=4, alpha=0.3)
ax.set_title("Projections of 3D lines")
# set limits
ax.set_xlim(-15, 15.5)
ax.set_ylim(-15.5, 15)
plt.show();
(Answer to question 1) To plot the intersections between the surface and the specified planes (y=-20, and y=20), one need to find what Y[?]=-20 and 20. By inspection, I found that Y[100]=20, Y[20]=-20.
The relevant code to plot the lines of intersection:
# By inspection, Y[100]=20, Y[20]=-20
ax.plot3D(X[100], Y[100], Z[100], color='red', lw=6) # line-1 at y=20
ax.plot3D(X[20], Y[20], Z[20], color='green', lw=6) # line-2 at y=-20
# Project them on Z=-100 plane
ax.plot3D(X[100], Y[100], -100, color='red', lw=3) # projection of Line-1
ax.plot3D(X[20], Y[20], -100, color='green', lw=3) # projection of Line-2
The output plot:
(Answer to question 2) To get better plot with the wireframe standout from the surface plot. The surface plot must be partially transparent, which is achieved by setting option alpha=0.6. The relevant code follows.
Z1 = Z-Z.min()
Z1 = Z1/Z.max()
Xall = zoom(X,3)
Yall = zoom(Y,3)
Zz = zoom(Z1, 3)
surf = ax.plot_surface(Xall, Yall, Zz, rstride=10, cstride=10,
facecolors = cm.jet(Zz/np.amax(Zz)),
linewidth=0, antialiased=True,
alpha= 0.6)
# Wireframe
ax.plot_wireframe(X, Y, Z, rstride=5, cstride=5, color='black', alpha=1, lw=0.8)
The plot is:
I'm fairly new to scatter plots and python in general. I am trying to plot a third variable against an x and a y, however, I'm not quite sure how to about specifying that argument? So I would have X values which are ints, y values which are also ints and then on the graph itself I want the model scores to show. Is there any way to do this sort of thing?
Thank you.
You can use color to plot a third value. Here is a very minimal example :
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import numpy as np
x = np.random.rand(100)
y = np.random.rand(100)
z = np.random.rand(100)
plt.scatter(x,y, c=z, s=5, cmap=cm.hsv)
cbar= plt.colorbar()
plt.show()
Edit
You could also use the size of markers, their transparency, hue or rgb values to depict even more information. Here is an example with marker size, alpha level and color on a perceptually uniform colormap.
import matplotlib.pyplot as plt
import numpy as np
import matplotlib.colors as colors
import matplotlib.cm as cmx
x = np.random.rand(100)
y = np.random.rand(100)
z = np.random.rand(100)
t = np.random.rand(100)
w = np.random.rand(100)
fig, ax = plt.subplots(1, 1)
cmap = plt.get_cmap('plasma')
cNorm = colors.Normalize(vmin=0, vmax=max(z))
scalarMap = cmx.ScalarMappable(norm=cNorm, cmap=cmap)
for i in range(100):
ax.scatter(x[i],y[i], c=scalarMap.to_rgba(z[i]), s=t[i]*100, cmap=cmx.plasma, alpha=w[i], edgecolor='none')
scalarMap.set_array([])
fig.colorbar(scalarMap,ax=ax)
for a in [0.1, 0.5, 0.9]:
ax.scatter([], [], c='k', alpha=0.5, s=a*100, label=str(a), edgecolors='none')
l1 = ax.legend(scatterpoints=1, frameon=True, loc='lower left' ,markerscale=1)
for b in [0.25, 0.5, 0.75]:
ax.scatter([], [], c='k', alpha=b, s=50, label=str(b), edgecolors='none')
ax.legend(scatterpoints=1, frameon=True, loc='lower right' ,markerscale=1)
fig.show()
At face value, that question doesn't really make sense because a conventional scatterplot has only two axes, and of course you can't plot points with three dimensions (x, y and accuracy).
However, there are alternative ways to do so.
Use colours
import numpy as np
from matplotlib import pyplot as plt
x = np.random.rand(200)
y = np.random.rand(200)
fig, ax = plt.subplots(figsize=(5, 5))
ax.scatter(x, y, c=(x + y), cmap='RdPu')
scatter takes a c argument, which can be a numeric value, as well as a cmap argument, which can be a string referencing a colormap.
The colormap object translates the numbers provided in c into points along a colour mapping, which you can think of as a gradient bar.
Use 3D axes
from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure(figsize=(5, 5))
ax = Axes3D(fig)
ax.scatter(x, y, (x + y))
This turns your 3rd dimension, accuracy, into an ordinary spatial dimension.
Use size of the markers
Very similar to the color option in the first part, you can change the size of the scatter markers (given you have some idea about the scale of the values). So based on the first example, you can also do;
import numpy as np
from matplotlib import pyplot as plt
x = np.random.rand(200)
y = np.random.rand(200)
fig, ax = plt.subplots(figsize=(5, 5))
ax.scatter(x, y, c='k', s=5*(x + y), cmap='RdPu')
scatter takes also the s argument, that changes the size of the markers.
I want to create a plot which looks similar to: https://www.shutterstock.com/de/image-vector/vector-abstract-3d-wave-wireframe-surrounding-445020520
I use something like the wire frame demo with a different background (best case a blue color gradient):
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt
from matplotlib import cm
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d', axisbg='blue')
# Grab some test data, e.g.gcreate noisy sea surface data
X, Y, Z = axes3d.get_test_data(0.05)
# Plot a basic white wireframe for the surface
ax.plot_wireframe(X, Y, Z, rstride=10, cstride=10, color='white')
plt.show()
What would you take to make this more look like the example?
Cheers Bene
Here is one (out of many) options. Using a surface plot in the background of the wireframe can give the plot some shading and make it more look like the picture.
import numpy as np
from mpl_toolkits.mplot3d import axes3d
import matplotlib.pyplot as plt
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d', facecolor='#1a5a98')
plt.subplots_adjust(0,0,1,1)
fig.patch.set_color('#1a5a98')
#Generate some data
x = np.linspace(0,500,501)/(9*np.pi)
X,Y = np.meshgrid(x,x)
f = lambda x,y: np.sin(x+y**0.2)*np.cos(y-x**0.4)
Z = f(X,Y)
# plot surface, for nice shading
ax.plot_surface(X, Y, Z, rstride=10, cstride=10)
# plot wireframe for white lines on top
ax.plot_wireframe(X, Y, Z, rstride=10, cstride=10, color='white', linewidth=0.5)
#set axis off
ax.axis("off")
ax.set_xlim(6.42,11.17)
ax.set_ylim(6.42,11.17)
ax.set_zlim(-6,6)
ax.view_init(elev=25, azim=-88)
plt.show()